Bay Area Decision Sciences Summit
This content is password protected. To view it please enter your password below:
The post Protected: The Next “Next Big Thing”: Agentic AI’s Opportunities and Risks appeared first on UC Berkeley Sutardja Center.
As an undergraduate student at UC Berkeley, Shuo Chen triple majored in Business Administration, Economics, and Rhetoric, minored in Public Policy, and earned the SCET Certificate in Entrepreneurship & Technology. Today, she is a General Partner at IOVC, investing in early-stage enterprise and SaaS startups. Additionally, she is an instructor at UC Berkeley and Stanford University, teaching topics around entrepreneurship and innovation.
At UC Berkeley, Chen took advantage of every opportunity to learn. Originally, she was uncertain about what she wanted to study. She chose UC Berkeley for its unique flexibility in selecting a major, and she especially appreciated how the breadth requirements encouraged students to broaden their horizons. Chen explored a plethora of academic disciplines, and she ultimately decided to major in three entirely different disciplines: Business Administration, Economics, and Rhetoric. Chen also quickly discovered her passion for entrepreneurship and immersed herself in the startup ecosystem on campus.
During her undergraduate years, Chen founded her first startup, Cashify. She realized she did not know much about managing her money, especially as an international student from Canada. To address this issue, she developed a curriculum around personal finance and money management, which later became the foundation of her startup. To develop the curriculum, Chen partnered with Haas faculty and worked to establish a DeCal course for undergraduate students, the very first personal finance course offered at Haas. The course was immediately a hit—though they did not market the class publicly, they received over 100 eager students on the first day of class. The overwhelming demand prompted Chen to make this content more readily available to different schools and companies. As a result of high demand, she founded Cashify, a social enterprise at the intersection of Fintech and EdTech, designed to provide financial literacy and money management training content and software for higher education and financial institutions. Throughout her undergraduate years, the business steadily evolved into a larger initiative that brought together a large group of contributors. Eventually, the entire team was hired in finance and Chen began her career in investment banking.
After graduating from UC Berkeley in three years, Chen worked in investment banking at Goldman Sachs before transitioning to a career in venture capital. A lifelong learner, Chen appreciated how being a venture investor seamlessly integrated continuous learning into her career.
“It’s an incredibly humbling process. My job is to work with some of the smartest people in the world trying to solve some of the world’s biggest problems, and my job is to learn from them and to identify those who are most likely to make that more positive future come true and to help those founders make that happen.”
Chen, who has had experience on both sides of the table as an entrepreneur and as an investor, notes that these roles differ significantly in skillset and scope. As an investor, Chen must support a range of founders and must adeptly identify patterns across multiple startups.
“Entrepreneurs are ultimately the ones who are building day to day. Our role as investors is to support them in the way they want to be supported. But because we’re not so involved in the day-to-day of operating one single company solving one specific problem, it sometimes is easier to pick up patterns for how you can do something more effectively. You can learn from one and apply to another company facing a similar challenge, even if they’re building a completely different business model.”
Equally important, Chen emphasized that maintaining aligned interests between founders and investors is essential for fostering a synergistic relationship, the cornerstone of a successful partnership.
“I used to think that investors were on the other side of the table because I was a founder myself, so I always thought that investors were not necessarily preparing for my best interest because they were on the other side of the table. And it took a lot of learning for me to understand why. There are incentives that are driving different behavior between founders and investors, for sure, but the best investors are those that are aligned with the interest of the founder from the get-go.”
When speaking with entrepreneurs, Chen notes that two key factors: the first is metrics around the business itself, such as efficiency around capital, tech, time, and growth, and the second is that the entrepreneur has a deep passion for the problem they are trying to solve.
Additionally, Chen explained a cultural unfavorability for fractional founders, individuals who take on the responsibility of a founder but on a part-time basis while maintaining a full-time role. She noted that, “I felt like that was quite taboo. People didn’t like the fact that you were doing multiple things.” However, Chen found herself continuously pursuing part-time projects to address the challenges she faced in her full-time roles as a student and later as an investment banker at Goldman Sachs.
“I always studied multiple things, and I had a breadth of interest. I was focused on trying to solve problems, but I felt like I was better if I were able to try to tackle the same problem from different angles at the same time.”
“There is a fine line to be drawn—if you’re doing multiple things that are completely unrelated, that indicates a lack of focus. But if you’re doing multiple things that contribute towards the goal, it can be a good opportunity for a founder to change the way that things are done because they can bring in different perspectives.”
Chen shared that her work is inspired by her struggles as a founder, stating that as an investor, she aims to help alleviate some of those challenges for others. For example, Chen faced pressure to sell her company to large companies and institutions, and she experienced a lack of available resources that would have deepened her understanding of enterprise business models and made the process more accessible, as engaging some of the largest companies in the world is quite formidable as a young entrepreneur.
Now as an investor, she is focused on supporting fractional founders, a group underappreciated in the startup ecosystem. She established a fund to support founders who had experienced the problem firsthand, tested solutions, and gained traction before deciding to turn their part-time projects into full-time ventures. The perception that founders should dedicate all of their time to building their startups from inception is a hallmark of Silicon Valley startup culture; funders who indicate that they still have full-time roles are often screened out early by investors when applying to startup accelerator programs.
“I think that that’s a missed opportunity. I think there are a lot of really smart, fractional founders who are time-efficient and can build amazing products in the early days part-time. After they test all their hypotheses, they can build it full-time.”
Chen views her career as an ongoing journey of learning rather than a linear path with a clear destination. In reflecting on her journey, she has found fulfillment in being able to contribute to a founder’s success through collaboration.
“I’m happy to be the entrepreneur behind the entrepreneur and help founders be successful in their own right. I think the proudest I’ve ever been is witnessing founders that we’ve invested in achieving successful outcomes. An outcome isn’t just defined by exit, as in an IPO or an acquisition—it’s really any outcome that they would be incredibly proud of for themselves.”
As she continues on her career journey, Chen highlights the importance of leaving a positive mark on the world, and she has found her calling in increasing efficiency in the workplace.
“How do you leave the world in a better place than what I started? This is kind of cliche, but I think it’s also true. And everyone defines better in a slightly different way. What I consider as ‘better’ is defined as more efficient in our work lives. I would love to leave the world a better place by helping people have more enjoyable workflows during those hours when they’re working. How can I help people better access to software, better access to tools that allow them to automate the more boring aspects of their work, and increase the time spent engaging in more strategic and personal relationship-building.”
The post A Conversation with Venture Capitalist Shuo Chen appeared first on UC Berkeley Sutardja Center.
Between October 28 and 30, representatives of the SCET community attended TechCrunch Disrupt, a three-day conference held annually in San Francisco. The event featured keynotes from industry experts, roundtable discussions, and pitch sessions, offering attendees an unparalleled opportunity to learn about the most disruptive technologies and discover emerging startups. At TechCrunch Disrupt, four Berkeley startups—Optigenix, Moonshine, Uncracked, and Swipefund—pitched their companies during roundtable sessions. Each of these startups originated in classes at SCET. Below, we’ve shared some favorite moments from our time at TechCrunch Disrupt.
SCET sent a Berkeley Delegation to TechCrunch Disrupt consisting of Spring 2024 Collider Cup winners and selected SCET student leadership, faculty, and staff. All delegation members attended a professional development session before TechCrunch led by Dr. Christianna Taylor who discussed leadership as it relates to startup development. Taylor has extensive experience helping teams cultivate leadership excellence by implementing practical strategies that foster a culture of trust and innovation.
In addition to the leadership development session, the members of the Berkeley delegation received mentorship tips from SCET instructors Gert Christen and Naeem Zafar on how to take advantage of their time at TechCrunch. Zafar noted that exploring how technology interacts with real-world problems can expand one’s perspective, creating new opportunities for compelling companies. He encouraged attendees to attend at least six keynote presentations, visit 20 demos, and collect 30 contacts. TechCrunch Disrupt is an incredible opportunity to meet the next generation of founders and entrepreneurs creating groundbreaking companies. Most importantly, Zafar hopes to see student attendees apply their learnings to their projects upon returning to campus.
Course coordinator, Manan Bhagarava led the student delegation. He organized logistics and supported student attendees as they navigated the TechCrunch space for the first time. The event featured keynotes from industry experts, roundtable discussions, and pitch sessions, offering attendees an unparalleled opportunity to learn about the most disruptive technologies and discover emerging startups. Some keynote speakers were Alex Pall and Drew Taggart from the Chainsmokers, who discussed bringing value beyond their celebrity status and discussing their fund Mantis Capital.
SCET Decal Facilitator Maher Hasan described his time at TechCrunch Disrupt as transformative, sharing, “It was an incredible experience connecting with so many amazing people. Hearing positive feedback not only about my company but also about some of the projects my friends are working on really helped me build solid relationships.”
“The highlight was definitely getting advice from The Chainsmokers. I’ve been a huge fan of their music since middle school—hearing their songs on the radio and now getting personalized advice from them made things come full circle. Drew’s comment about how you will inevitably mess up resonated with me. It was a powerful reminder to be kinder to myself, especially coming from someone I’ve admired for so long.”
TechCrunch, which is held at the Moscone West Center in San Francisco, hosts over 10,000 attendees ranging from investors, startup founders, venture capitalists, students, and more. Through SCET’s partnership with TechCrunch, UC Berkeley was the only university to be an official sponsor at the conference.
Four SCET-founded companies had the incredible opportunity to showcase their startups at the TechCrunch roundtables. Both roundtable discussions hosted by SCET were at full capacity as investors, founders, industry leaders, and students witnessed firsthand what makes UC Berkeley the #1 university startup ecosystem in the world. Before the pitch, attendees learned about SCET’s Berkeley Method of Entrepreneurship, which emphasizes an inductive, journey-based approach to entrepreneurship education that actively engages students and challenges them to develop real-world entrepreneurial skills. After learning more about the SCET pedagogy, attendees heard four teams pitch.
“One of the core principles of SCET is to embrace the ‘innovation collider’ model—that is, to bring together people who are different to create new strengths and unique perspectives,” said SCET Chief Marketing Officer Keith McAleer. “Leading the Berkeley delegation at TechCrunch Disrupt was a great way to help innovative students share their ideas, create new connections, and also help SCET better connect with what is happening in the industry.”
The four teams that represented SCET are also teams that have deep roots in the Collider Cup. Optigenix, a company founded in ENGIN 183C – Challenge Lab: SportsTech, Entrepreneurship & the Future of Sports, competed but did not place in the 2023 competition. However, they returned as an alumni expo contestant in the Spring 2024 Collider Cup and won.
Pitching alongside Optigenix, was Moonshine, an AI video editor platform, the Collider Cup winners in Fall 2022. Co-founders Harsha Gundala and Ganesh Pimpale met each other in class and connected over a shared passion for video editing and startups. They are currently participating in Y Combinator’s Fall 2024 batch and, through the Roundtable, have had the opportunity to connect with other startup founders in similar industries.
The second roundtable session highlighted female founders, emphasizing UC Berkeley’s leadership in supporting female founders. Kate Sullivan, founder of UnCracked, a seafood sustainability company, was the winner of the Collider Cup in Spring 2023 and also participated as an alumni expo participant that same year. Uncracked, founded in ENGIN 183C – Challenge Lab: AltMeat: Product Design of Plant-Based Foods, initially focused on producing plant-based crab meat, but has since shifted to producing algae-based products. Since then, Kate has been able to connect with UC Berkeley and SCET alumni to further discuss investment opportunities. Three-time course coordinator Amy Jain and Shristi Dalal presented their startup Swipefund, founded in ENGIN 183 Startup Catalyst to an audience of eager investors and industry professionals.
TechCrunch Disrupt was a transformative experience for the Berkeley delegation—from engaging with industry trailblazers to learning about emerging technologies. The Berkeley Delegation is excited to take their learnings from TechCrunch Disrupt and apply them to their work, carrying forward the spirit of innovation and collaboration they witnessed on the global stage.
The post Collider Cup Winners Lead Berkeley Delegation at TechCrunch Disrupt appeared first on UC Berkeley Sutardja Center.
At the Sutardja Center for Entrepreneurship & Technology (SCET), we believe that the best way for students to learn entrepreneurship is through direct interaction with industry experts. Our instructors are accomplished entrepreneurs, executives, and leaders who have first-hand experience building businesses. This approach ensures that students learn practical skills from individuals who have tackled real-world challenges and understand what it takes to innovate in today’s fast-paced industries.
One such instructor is Sandy Diao, a seasoned industry leader with over a decade of experience leading growth teams at high-profile companies such as Meta, Pinterest, and Descript. With a deep background in growth marketing, Sandy is currently teaching two versions of the Growth Marketing course at SCET—one tailored for Berkeley students and another designed for professionals. Her classes focus on equipping students with practical skills in growth marketing, leveraging her background growing users from zero to hundreds of millions. Sandy’s teaching style is hands-on, blending her experience with real-time marketing insights to help students understand the core principles of growth marketing and how to apply them effectively.
Sandy’s unique journey into entrepreneurship and her passion for innovation were sparked during her time as a UC Berkeley SCET student. She credits a pivotal mentorship experience with reshaping her understanding of learning, growth, and what it means to innovate. In her classes, Sandy encourages students to adopt an entrepreneurial mindset by “doing things”—an approach that aligns with her own belief that knowledge is best gained through practice.
Here’s a Q&A with Sandy, where she shares insights into her teaching philosophy, the value of entrepreneurship skills, and her advice for students looking to make the most of their Berkeley experience.
Q: What inspired you to focus your teaching on entrepreneurship and technology innovation, and how has your industry and/or founder experience shaped your approach in the classroom?
Sandy Diao: My interest in entrepreneurship came in an unexpected way when I studied at Cal over a decade ago. I received a leadership scholarship as one of five recipients from Aaron Mendelson (who remains an amazing mentor and is a Cal Bear). This scholarship led me to take a private trip down to Sandhill Road, where I found myself sitting across from the early investors of Apple, Oracle, Facebook, Salesforce, and many of today’s giants. Watching these technologists piece together the future from ideas, experiences, and conviction changed everything for me. Aaron never taught me through a formal classroom setting, yet he profoundly taught me the most important lesson that I took away from school: real learning happens by doing things.
This lesson has held true for me throughout my roles as a growth executive serving some of the world’s fastest-growing companies and startups like Pinterest, Facebook, Instagram, Descript, and more. I leaned into the emerging field of growth, a role that didn’t even exist for new grads when I left school, by crafting opportunities through my bias for execution and getting things done.
That’s why I intend to teach my growth marketing class at SCET more like a startup workshop than a lecture hall. I’m building, launching, and learning alongside our students, and sharing the practical tools I wish I’d had when I graduated. I’m learning tons from SCET’s veteran teachers and staff, and I hope to show that doing gets to knowing, and knowing drives growing.
Q: Why do you think it is helpful for everyone, regardless of their field of study, to learn innovation and entrepreneurship skills?
Sandy Diao: From working at some of the fastest-growing startups, I’ve learned that entrepreneurship is fundamentally about creative problem-solving. It’s the scientific method in action: developing insights, forming hypotheses, and validating those hypotheses with experiment after experiment.
Whether you want to be a founder or not, entrepreneurial skills teach you how to solve problems with real-world constraints – time, people, resources. And that’s a superpower to be used in any career.
Q: What advice would you give to incoming freshmen or transfer students to help them maximize their potential and make the most of the opportunities available at Berkeley?
Sandy Diao: At UC Berkeley, you’ll get more than just a world-class education – you’ll find genuinely kind humans who’ll support you long after graduation.
Most students are great at diving into student organizations and internships. Looking back, I wish I had also invested more into relationships. The connections that I’ve maintained from Berkeley have had a huge impact on me personally and professionally. Those classmates you casually sit next to in your SCET class could become your future cofounder or even best friend. So grab that coffee or lunch and invest in the people around you. Ten years from now, you’ll be grateful for the ambitious friends who’ve known you since day one.
Q: What are some of the biggest challenges you see students facing when trying to apply entrepreneurial thinking to real-world problems, and how do you help them overcome those challenges?
Sandy Diao: I see the biggest challenge students face in entrepreneurial thinking as the gap between knowing and doing. For example, concepts like marketing funnels or SEO can make perfect sense in a classroom, but it can be hard to translate knowledge into doing because of the perception that doing something is harder than it actually is.
That’s why my approach in the classroom is to give students “training wheels” through live tutorials and hands-on exercises using the simplest tools. We start doing before overthinking. Most things are easier to get started with than they appear. The hardest part is often just beginning. I’ve found the best way to bridge the knowing-doing gap is by using training wheels – the easiest, simplest tools possible to overcome that initial cold start.
Through her teaching, Sandy Diao brings the dynamic world of growth marketing into the classroom, enabling students to learn by doing and developing the skills necessary for entrepreneurial success. Her classes at SCET provide a valuable opportunity for students to engage directly with industry practices, preparing them for the challenges of building and scaling innovative solutions in a rapidly changing world.
Aspiring and current entrepreneurs can connect with Sandy Diao on LinkedIn and can follow her writing on sandydiao.com.
The post SCET Instructor Highlight: Sandy Diao appeared first on UC Berkeley Sutardja Center.
Join us at Collider Cup XV, where innovative Berkeley minds converge for the university’s premier technology showcase and pitch competition.
Collider Cup XV is Berkeley SCET’s pitch competition and a showcase of Berkeley’s top student venture teams from the Fall 2024 semester. This event is the culmination of a semester’s hard work, where student teams pitch their ventures to win the sought-after Collider Cup. In addition to the competition, SCET will share insights about its upcoming courses and offer an opportunity for invaluable networking with free food after the event.
Throughout the semester, students from SCET’s Fall 2024 venture courses have honed their startup ideas. These courses, open to all majors, foster interdisciplinary teams focused on creating solutions for societal problems using cutting-edge technologies like AI, foodtech, and healthtech.
Learn more about the courses that have propelled these teams to the forefront:
For full course details, visit the SCET Courses Page.
After the pitches, stick around for complimentary food and networking in the Kvamme Atrium to connect with student innovators, faculty, and investors and embrace the spirit of entrepreneurship at UC Berkeley!
Arrive early and experience the second-ever SCET Student Alumni Collider Cup Expo! An hour before Collider Cup XIV kicks off, SCET alumni will showcase their ventures, competing for an interview with Pear VC. As a participant, you’ll play a crucial role in selecting the winning team, so come early to cast your vote and secure your seat for Collider Cup XIV.
→ Alumni (who have taken at least one SCET course before Fall 2024) are eligible to apply to join the expo. Apply by [date] for consideration.
The event will be held by Benecia Jude Jose, a fourth-year student in Public Health & Data Science, and Daniella Spero, a third-year student in Political Science and Public Policy. Jay Onda from Marubeni Ventures, Khalil Fuller from Pear VC, and Sibyl Chen from Berkeley SkyDeck will serve as the judge panel at the Collider Cup this fall. These judges bring a wealth of knowledge and experience to the table, providing students with invaluable feedback.
Collider Cup XV will award 1st, 2nd, 3rd, Most Innovative, and People’s Choice awards to competing teams. We’re thrilled to announce the following prizes from our exceptional partners:
For more information on the partners and detailed descriptions of the prizes, visit their respective websites:
The post Join us for Collider Cup XV — UC Berkeley’s Premier Technology Entrepreneurship Showcase appeared first on UC Berkeley Sutardja Center.
Nicolás Camhi joined the SCET community in 2018 after earning a scholarship from Pontificia Universidad Católica de Chile to participate in the Startup Semester program at UC Berkeley SCET. Fast forward to today, and he is currently the co-founderCamhi earned a degree in Industrial Engineering from Pontificia Universidad Católica de Chile. Coming from a family of entrepreneurs, he knew early on that he wanted to follow in their footsteps. Throughout his time at the university, he worked on numerous side projects, including designing a QR code payment method to help students skip the lunch line in Chilean schools.
After winning a scholarship through the LeBridge Startup Fellows Program at UC Berkeley, he traveled in a group of thirty students from around the world to SCET. Throughout his Startup Semester, Camhi worked as a product manager on a tech startup called Urban Radar. His time at UC Berkeley deeply inspired Camhi, and he decided to return to the Bay Area immediately after wrapping up his career as an industrial engineer to start his own business.
Camhi partnered with Matías Pérez Pefaur and Diego Chahuán Luhrs, both Civil Engineers at Pontificia Universidad Católica de Chile. Once Nico identified an opportunity to start a business, he reached out to Pefaur for his technical expertise. The two began developing debt collection software. After facing difficulties with the initial product, Diego joined the team, overseeing technology, scalability, and operations. Matías leads AI and product development, and Nico handles executive duties, sales, investor relations, and team management.
The team decided to pivot after realizing that he could use AI to collect debt through WhatsApp. In early March of this year, they switched gears to build an AI-based business communication platform based on customer feedback, specifically in Latin American customers. It was here that Vambe was born — a solution designed to help small and medium-sized businesses scale their sales processes in Latin America, a region where the conversational economy is dominant. Often, companies of this size lack the bandwidth to manage all customer engagement interactions, particularly messages through Instagram, Facebook, and WhatsApp, without incurring significant costs. Vambe is a solution utilizing an AI platform to drive this traffic.
Raised by a family of entrepreneurs, Camhi recognized this issue, found inspiration, and took action. He says, “I’ve been in the shoes of my customers since I was very young. At Vambe, we serve small to medium-sized businesses, and my family could have been my potential customer.” He further noted that it is these entrepreneurs who are driving the economy.
“I am really inspired by them. Now, being in a position to help these entrepreneurs drive their businesses more efficiently and improve their sales is something that motivates me.” and CEO at Vambe (SkyDeck Batch 18), a company providing a hyper-personalized conversational AI platform for small and medium-sized businesses to increase their sales capacity and boost closing rates. Vambe has achieved the highest annual recurring revenue (ARR) of all SkyDeck Batch 18 companies, scaling from 24k to 500k in ARR in just five months, which is an 80.2% average month-over-month growth. Currently, the team is raising their first seed round. We connected with co-founder Nicolas Camhi to hear about his reflections on his entrepreneurial journey and his vision for the future of Vambe.
Since pivoting, the team has overcome numerous technical challenges, achievements that would help them earn their spot as the highest annual recurring revenue (ARR) of all SkyDeck Batch 18 companies. In a short time, they have designed an intuitive platform for customers who do not have an extensive technical background. Additionally, they have been working on advancing AI capabilities without requiring prompt engineering expertise, and they have been scaling customer communications and simultaneously orchestrating multiple AI flows.
The company is currently focused on streamlining the customer onboarding process, developing self-service capabilities, and expanding integrations with other SaaS platforms and customer relationship management platforms.
When asked what he loves most about what he does, Camhi replied that tackling difficult problems with an intelligent and collaborative team is the most rewarding aspect of his work.
“When I’m there with the team thinking about how we’re going to fix a problem, improve a bottleneck in our processes, deliver our solution better, improve our product to better serve our customers and add more value. That’s the most fun part of it by far – taking on big challenges with a wonderful team.”
When asked about the qualities shared by all successful entrepreneurs, Camhi emphasizes the importance of resilience.
“Entrepreneurship is a roller coaster of emotions – it’s real and extreme. One day, you could be at the top, and at the end of the day, you could be down. There are a large number of different problems coming at you that you don’t know how to solve right away, and you don’t have the experience to help you solve them. It’s intense in every sense, and being resilient and flexible is a requirement to adapt to all of these situations.”
Additionally, he expressed just how critical a good work ethic is to achieving entrepreneurial success. He says, “You have to be really obsessed with what you’re doing. You need to hustle and put your everything into it.”
Camhi’s advice to all prospective entrepreneurs is to just jump in. He says, “If you are really sure that you want to do this, go for it. I’m just twenty-six and have been doing entrepreneurship since I was eighteen years old. Most of the people I meet have this approach of gaining experience in the corporate workplace before pursuing entrepreneurship. I don’t think that’s true – I would tell students and young entrepreneurs to chase after their dreams if they are passionate about an opportunity.”
In the next couple of years, Camhi’s goal is to focus on expanding in Latin America and eventually dominate that space. Ultimately, the team hopes to go global, bringing Vambe to businesses in the US and Europe. Camhi says, “I would like to see Vambe supporting every entrepreneur, providing our platform to make AI applications real and accessible to all entrepreneurs who want to drive their businesses.”
In reflection on his entrepreneurial journey so far, Camhi noted that he has come a long way. “When I was twenty years old, I came here to Berkeley as a student chasing a dream of one day becoming an entrepreneur. Almost six years ago, I came to SkyDeck and remember seeing how entrepreneurs work. And now, I’m back here again, but as a founder. I’m speaking at our demo day in front of over 1,400 investors and receiving students. It’s super amazing for me.”
Camhi hopes to build and lead a company that will encourage others to become the best version of themselves.
“What I love to do is to inspire people to believe in themselves. It’s what I’ve been doing. I would love to build a company and a group of people to help them make the most of their potential. That’s what I try to do with my customers — helping other people reach their maximum potential, as well as their businesses. Our slogan is ‘AI for all, growth for all.’”
In the long term, the Vambe team seeks to make AI accessible to businesses of all sizes and functions. Eventually, the founders see the company evolving into a primarily data-driven company that improves the commerce experience orchestrated with artificial intelligence.
The post Inside Vambe: A Conversation with Founder Nicolás Camhi appeared first on UC Berkeley Sutardja Center.
This content is password protected. To view it please enter your password below:
The post Protected: Joo Ae Chu, SCET’s Academic Program Manager, Receives UC Berkeley’s 2024 Advising Innovation Award appeared first on UC Berkeley Sutardja Center.
SCET has renewed their partnership with Plaksha University, a technology university in the Chandigarh area of India. The newest agreement strengthens a partnership of over three years between SCET and Plaksha which developed an academic minor in technology entrepreneurship and design and delivery of the entrepreneurship curriculum for the Technology Leaders Fellowship Program (TLF).
Since Plaksha began in 2019, they have been inspired in their curriculum development by SCET’s Berkeley Method of Entrepreneurship, which engages students in an immersive, hands-on experience with innovation. SCET faculty have taught and advised students in the innovation program co-created with Plaksha.
The new one-year postgraduate program will nurture exceptional innovators to develop entrepreneurial thinking and gain real-world tech leadership experience. Berkeley mentors, such as Alexander Fred-Ojala and Ken Singer, coach students through challenges similar to SCET’s Collider Cup. Following the partnership renewal, TLF will undergo a hiatus this academic year and will relaunch as a one-year Master’s program in 2025.
Berkeley’s groundbreaking entrepreneurship methods are a major pull factor for students considering Plaksha. Mark Searle, the Head of Mentoring Excellence at SCET and a course coordinator at Plaksha, stated that many students choose to enroll at Plaksha because of the opportunity to connect with eminent universities in India and overseas.
In line with the mission of reimagining technology education, Plaksha is making the undergraduate curriculum interdisciplinary. This August, they launched a new minor in technology and innovation for the BTech program with guidance and inputs from Berkeley. Susan Giesecke, SCET’s Director of Global Engagement, explained that the universities are aiming to co-create a learning opportunity that can function similarly to the TLF program and inform the next generation of Indian technology leaders.
The post Renewed Plaksha University partnership expands BMoE through revamped curriculum appeared first on UC Berkeley Sutardja Center.
Your cell phone has smokestack emissions. So too does your electric vehicle. The simple reason for this is that, here in the US, only 3.6% of energy supply in 2023 came from renewable sources such as wind, solar, hydroelectric and geothermal. Fossil fuels remain our predominant way of generating electricity.
The picture is equally bleak on a worldwide basis, with fossil fuels meeting the bulk of our energy demands, turning every electrically powered device, from cell phone, to electric vehicle, to data center server, into an exhaust-spewing challenge to Mother Nature.
In our increasingly digitized world, the needs from computation, and artificial intelligence in particular, are now creating a profound new climate stress. As the use of AI spreads into all aspects of human life and business, it extends ever more its demands for power and water.
Just how much power and water are we talking about? AI-driven data center power consumption is slated to soon reach 8.4 TWh (Chien 2023) which “is 3.25 gigatons of CO2, the equivalent of 5 billion U.S. cross-country flights”. Yale’s School of the Environment found that “Data centers’ electricity consumption in 2026 is projected to reach 1,000 terawatts, roughly Japan’s total consumption” (Berreby 2024), while researchers at UC Riverside found that the global AI demand for water in 2027 would be 4.2-6.6 billion cubic meters, roughly equivalent to the consumption of “half of the United Kingdom” (Li et al. 2023). Ouch.
To address the accelerating demands of AI’s energy consumption, the ideal solution would be to transition to 100% renewable energy, but this goal is currently distant. A more feasible approach is the syncretic one, combining specialized AI hardware and software, innovative data center designs, and the implementation of comprehensive AI policies, including regulation. This discussion will outline current strategies for reducing AI’s energy demands, with many solutions derived from software technology, thus enhancing their accessibility to AI practitioners.
To address the heat generated by AI computations in data centers, which necessitates significant cooling energy, multiple different cooling methods can be employed. Free air cooling, which uses outdoor air to cool indoor environments, is highly efficient and uses minimal water but only works in cooler climates. Evaporative (adiabatic) cooling also provides efficient cooling with low power and water usage. Some recent designs utilize submersion cooling, where hardware is immersed in a dielectric fluid that transfers heat without conducting electricity, thus eliminating the need for traditional air conditioning. Conversely, mechanical air conditioning is least effective due to high power and water costs.
While nuclear power via small modular reactors (SMRs) is sometimes proposed as a cleaner energy solution for data centers, their deployment will take years, and they face similar safety, waste, and economic concerns as larger reactors. Instead, focusing on renewable energy sources and storage solutions, such as solar power, may offer more immediate benefits for meeting data center energy needs.
Since the mid-19th century, internal combustion engines have evolved into highly specialized devices for various applications, from lawnmowers to jumbo jets. Today, AI hardware is undergoing a similar evolution, with specialized, high-performance processors replacing less efficient general-purpose CPUs. Google’s introduction of the tensor processing unit (TPU) in 2015 marked a bellwether advance in custom-designed AI hardware. NVIDIA’s GPUs, which excel in parallel processing for deep learning and large-scale data operations, have driven substantial growth in both sales and stock valuation. As AI demand increases, we can anticipate a proliferation of vendors offering increasingly specialized processors that deliver superior computation speeds at lower energy costs.
Recent examples of this inevitable wave include Groq, a company building a language processing unit (LPU). Groq claims its custom hardware runs generative AI models similar to those from OpenAI “at 10x the speed and one-tenth the energy”. Also included in the list is Cerebras Systems’ wafer-scale chip, which “runs 20 times faster than NVIDIA GPUs”. Then there’s Etched’s Sohu ASIC, which burns the transformer architecture directly into the silicon, so “can run AI models an order of magnitude faster and cheaper than GPUs”; SiMa.ai which claims “10x performance”; and of course Amazon’s Trainium, Graviton and Inferentia processors.
The future of chip innovation may lie in biomimetic design, inspired by nature’s energy-efficient intelligence. Technologies like those developed by FinalSpark are expected to contribute to this trend. However, access to specialized processors is likely to remain a competitive landscape, with smaller companies facing particular challenges.
Data center cooling equipment includes chillers, pumps and cooling towers. Can this equipment be run in an optimal fashion in order to maximize cooling while minimizing energy? In 2016, engineers at Google did just that, implementing a neural network (Evans & Gao 2016) featuring five hidden layers with 50 nodes per layer, and 19 discrete input variables, including data from the cooling equipment and weather conditions outdoors. Trained on 2 years’ worth of operating data, this neural network succeeded in reducing the energy used for cooling Google’s data centers by a whopping 40%. (Despite this, Google’s greenhouse gas emissions have skyrocketed 48% in the past 5 years thanks to the relentless demands of AI.)
In addition to using software to orchestrate the cooling machinery in a data center, the same can be done with the AI applications running there. By optimizing what gets run (if at all), where it’s run and when it’s run, data centers can achieve substantial energy savings on their AI workloads. (As a simple example, imagine moving an AI workload from the afternoon to the early morning hours and saving 10% of the energy right off the bat.).
Orchestration of AI falls into two categories: orchestrating the AI training process, and orchestrating AI at inference (runtime). There are a number of different approaches being taken in the orchestration of AI training. Two promising tacks are power-capping (McDonald et al. 2022) and training performance estimation (TPE; Frey et al. 2022). In the former, standard NVIDIA utilities were used to cap the power budget available for training a BERT language model. Though the power cap led to a longer time-to-train, the resulting energy savings were material, with “a 150W bound on power utilization [leading] to an average 13.7% decrease in energy usage and 6.8% increase in training time.”
TPE is based on the principle of early stopping during AI training. Instead of training every model and hyperparameter configuration to full convergence over 100 epochs, which incurs significant energy costs, networks might be trained for only 10-20 epochs. At this stage, a snapshot is taken to compare performance with other models, allowing for the elimination of non-optimal configurations and retention of the most promising ones. “By predicting the final, converged model performance from only a few initial epochs of training, early stopping [of slow-converging models] saves energy without a significant drop in performance.” The authors note that “In this way, 80-90% energy savings are achieved simply by performing HPO [hyperparameter optimization] without training to convergence.”
Approximately 80% of AI’s workload involves inference, making its optimization crucial for energy reduction. An illustrative example is CLOVER (Li et al., 2023), which achieves energy savings through two key optimizations: GPU resource partitioning and mixed-quality models. GPU partitioning enhances efficiency by orchestrating resource utilization at the GPU level. Mixed-quality models refers to the availability of multiple models with different qualities in accuracy and resource needs. “Creating a mixture of model variants (i.e., a mixture of low- and high-quality models) provides an opportunity for significant reduction in the carbon footprint” by allowing the best model variant to be orchestrated at runtime, trading off accuracy against carbon savings. CLOVER’s mixed-quality inference services combined with GPU partitioning has proven highly effective, yielding “over 75% carbon emission savings across all applications with minimal accuracy degradation (2-4%)”.
Orchestration from a portfolio of mixed-quality models brings tremendous promise. Imagine intelligently trading off energy versus accuracy at runtime based on real-time requirements. As a further example, it’s been shown that “the answers generated by [the smaller] GPT-Neo 1.3B have similar quality of answers generated by [the larger] GPT-J 6B but GPT-Neo 1.3B only consumes 27% of the energy” and 20% as much disk (Everman et al. 2023). Yet another impactful approach to orchestration by cascading mixed-quality LLMs was shown in FrugalGPT (Chen et al. 2023). A key technique in FrugalGPT was to use a cascaded architecture to avoid querying high-resource-demand GPT-4 as long as lower-resource-demand GPT-J or J1-L were able to produce high-quality answers. “FrugalGPT can match the performance of the best individual LLM (e.g. GPT-4) with up to 98% cost reduction or improve the accuracy over GPT-4 by 4% with the same cost”.
Also notable in orchestrating machine learning inference is Kairos (Li et al. 2023). Kairos is a runtime framework that enhances machine learning inference by optimizing query throughput within quality and cost constraints. It achieves this by pooling diverse compute resources and dynamically allocating queries across that fabric for maximum efficiency. By leveraging similarities in top configurations, Kairos selects the most effective one without online evaluation. This approach can double the throughput of homogeneous solutions and outperform state-of-the-art methods by up to 70%.
From an innovation standpoint, a market opportunity exists for providing optimized AI orchestration to the data center. Evidence of the importance of AI orchestration may be seen in NVIDIA’s recent acquisition of Run:ai, a supplier of workload management and orchestration software.
Reducing the complexity of mathematical operations is a key strategy for decreasing AI’s computational load and energy consumption. AI typically relies on 32-bit precision numbers within multi-dimensional matrices and dense neural networks. Transformer-based models like ChatGPT also use tokens in their processing. By minimizing the size and complexity of numbers, matrices, networks and tokens, significant computational savings can be achieved with minimal loss of accuracy.
Quantization, the process of reducing numerical precision, is central to this approach. It involves representing neural network parameters with lower-precision data types, such as 8-bit integers instead of 32-bit floating point numbers. This reduces memory usage and computational costs, particularly for operations like matrix multiplications (MatMul). Quantization can be applied in two ways: post-training quantization (PTQ), which rounds existing networks to lower precision, and quantization-aware training (QAT), which trains networks directly using low-precision numbers.
Recent work with OneBit (Xu et al. 2024), BitNet (QAT: Wang et al. 2023) and BiLLM (PTQ: Huang et al. 2024) have shown the efficacy – delivering accuracy while reducing memory and energy footprint – of reduced bit-width approaches. BiLLM, for example, approximated most numbers with a single bit, but utilized 2 bits for salient weights (hence average bit-widths > 1). With overall bit-widths of around 1.1, BiLLM was able to deliver consistently low perplexity scores despite its lower memory and energy costs.
Reinforcing the potential for 1-bit LLM variants is BitNet b1.58 (Ma et al. 2024), “where every parameter is ternary, taking on values of {-1, 0, 1}.” The additional value of 0 was injected into the original 1-bit BitNet, resulting in 1.58 bits in the binary system. BitNet b1.58 “requires almost no multiplication operations for matrix multiplication and can be highly optimized. Additionally, it has the same energy consumption as the original 1-bit BitNet and is much more efficient in terms of memory consumption, throughput and latency compared to FP16 LLM baselines”. BitNet b1.58 was found to “match full precision LLaMA LLM at 3B model size in terms of perplexity, while being 2.71 times faster and using 3.55 times less GPU memory. In particular, BitNet b1.58 with a 3.9B model size is 2.4 times faster, consumes 3.32 times less memory, but performs significantly better than LLaMA LLM 3B.”
Ternary neural networks (Alemdar et al. 2017) that constrain weights and activations to {−1, 0, 1} have proven to be very efficient (Liu et al. 2023, Zhu et al. 2024) because of their reduced use of memory and ability to eliminate expensive MatMul operations altogether, requiring simple addition and subtraction only. Low-bit LLMs further lend themselves to implementation in (high-performance) hardware, with native representation of each parameter as -1, 0, or 1, and simple addition or subtraction of values to avoid multiplication. Indeed, the Zhu work achieved “brain-like efficiency”, processing billion-parameter scale models at 13W of energy (lightbulb-level!), all via a custom FPGA built to exploit the lightweight mathematical operations.
Low-Rank Adaptation (LoRA; Hu et al. 2021) is a technique designed to fine-tune large pre-trained models efficiently by focusing on a low-rank (lower-dimensionality) approximation of the model’s weight matrices. Instead of updating all the model parameters, LoRA introduces additional low-rank matrices that capture essential adaptations while keeping the original weights mostly unchanged. This approach reduces computational costs and storage requirements, making it feasible to adapt large models to specific tasks with limited resources. “LoRA can reduce the number of trainable parameters by 10,000 times and the GPU memory requirement by 3 times.” In addition, LORA provides better model quality “despite having fewer trainable parameters, a higher training throughput, and, unlike adapters, no additional inference latency”.
Neural network pruning achieves computational efficiency via the technique of reducing the size of a trained neural network by removing unnecessary connections (weights) or even entire neurons, thus making the network smaller, faster, and more efficient without significantly sacrificing performance. The concept was first introduced in the paper “Optimal Brain Damage” (Le Cun et al. 1989), and has been much advanced (Han et al. 2015) in the recent past. The efficiencies reported in Han et al.’s work were substantial: “On the ImageNet dataset, our method reduced the number of parameters of AlexNet by a factor of 9×, from 61 million to 6.7 million, without incurring accuracy loss. Similar experiments with VGG-16 found that the total number of parameters can be reduced by 13×, from 138 million to 10.3 million, again with no loss of accuracy.”
Knowledge distillation is a technique for reducing neural network size by transferring generalized knowledge from a larger “teacher” model to a smaller “student” model, perhaps similar to the Hoskins Effect in virology. This process involves distilling the teacher model’s probability distributions into the student model, which results in a more compact network that maintains high performance at lower resource costs. Knowledge distillation has proven effective in tasks such as image identification (Beyer et al., 2021), pedestrian detection (Xu et al., 2024), and small language models. For instance, NVIDIA and Mistral AI’s Mistral-NeMo-Minitron 8B achieved superior accuracy compared to other models by combining neural network pruning and knowledge distillation, despite using orders of magnitude fewer tokens.
Small language models (SLMs) also offer a method to reduce computational load and energy consumption. While SLMs are often discussed in the context of on-device applications, such as those from Microsoft and Apple, they also decrease computational and energy demands in data center environments. SLMs are characterized by smaller datasets, fewer parameters, and simpler architectures. These models are designed for low-resource settings, requiring less energy for both inference and training. Research (Schick & Schütze 2020) indicates that SLMs can achieve performance comparable to GPT-3 while having orders of magnitude fewer parameters.
Another notable optimization approach is SPROUT (Li et al., 2024), which reduces transformer math and carbon impact by decreasing the number of tokens used in language generation. SPROUT’s key insight is that the carbon footprint of LLM inference depends on both model size and token count. It employs a “generation directives” (similar to compiler directives) mechanism to adjust autoregressive inference iterations, achieving over 40% carbon savings without compromising output quality.
One last method for reducing AI’s computational load that brings promise is neuro-symbolic AI (NSAI; Susskind et al. 2021). NSAI integrates neural networks with symbolic reasoning, combining the strengths of both: neural networks excel at pattern recognition from large data sets, while symbolic reasoning facilitates logic-based inference. This integration aims to overcome the energy demands of neural networks and the rigidity of symbolic systems, creating more robust and adaptable AI. Research indicates that NSAI can achieve high accuracy with as little as 10% of the training data, potentially representing a pathway to sustainable AI.
It should be noted that pushing AI computation to the edge, e.g., onto your mobile device, does have the effect of reducing the 40% of data center energy that’s currently spent on cooling. Energy and carbon impacts are however still incurred from charging your mobile device.
The battle for AI chip supremacy is being fought equally on the silicon as well as on the software framework that’s built atop the silicon. These frameworks increasingly provide native support for the sorts of math optimizations that have been described in this article. Pruning, for example, is one of the core optimization techniques built into TensorFlow MOT.
NVIDIA competitor, AMD, has been aggressively accreting software framework technology via the acquisition of companies such as Mipsology, Silo.ai and Nod.ai. All in aid of countering the significant advantages brought to NVIDIA’s hardware by its extensive software technology, including its CUDA parallel programming platform and NIM (NVIDIA Inference Microservices) framework.
In NVIDIA’s recent work published together with Hugging Face, the full impact of turning NIM on was seen in the 3x improvement in tokens/second performance. Note the functionality embedded within NIM.
Modern software applications, from lightweight mobile apps like Instagram to complex systems such as automobile operating systems, can encompass 1 million to 100 million lines of code. This underscores the necessity of integrating energy-awareness into software design from the outset, treating it as an architectural attribute alongside scalability and latency. Neglecting this early integration will otherwise result in a challenging / insuperable retrofitting process for energy-efficiency at some later point, in what will eventually become a large, legacy application.
Key architectural strategies include simplifying code through pruning layers and nodes, reducing instruction counts, training with minimal data as in SLMs, and employing techniques like RAG and p-tuning to minimize training overhead. Additionally, incorporating drift tolerance, zero-shot and transfer learning, optimizing job schedules, and carefully selecting cloud computing resources are essential practices.
Of salient importance is also the requirement of measuring the climate impacts of AI models. Per the old saw, you can’t improve what you can’t measure. The tools for monitoring the energy footprints from AI are many, readily supplied by cloud vendors such as Amazon, Google, Microsoft and NVIDIA. As well, there are multiple third-party solutions available from the likes of Carbontracker, Cloud Carbon, PowerAPI, CodeCarbon, ML Commons and ML CO2 Impact.
Watt’s Up? Policies!
“Into the corner, broom! broom! Be gone!” from Goethe’s The Sorcerer’s Apprentice
AI is becoming ever more ubiquitous, leading to ever larger demands on our straining power grid. Despite all of the measures that can be taken to dampen the power and environmental impacts of AI, such as the methods described here, AI technology is fighting a losing battle with itself. The International Energy Agency recently found that “The combination of rapidly growing size of models and computing demand are likely to outpace strong energy efficiency improvements, resulting in a net growth in total AI-related energy use in the coming years.” This conclusion mirrored one that was reached by researchers from MIT (Thompson et al. 2022), which stated “that progress across a wide variety of applications is strongly reliant on increases in computing power. Extrapolating forward this reliance reveals that progress along current lines is rapidly becoming economically, technically, and environmentally unsustainable”.
What can we do? The answer lies in governing our actions through policies at three levels: as individuals, as corporations, and as polities. Reducing AI energy demand stands as both a moral imperative and a sound business practice.
To mitigate AI’s climate impact, regulatory measures are inevitable. Following the 1973 oil crisis, the precedent set by 1975’s Corporate Average Fuel Economy (CAFE) standards, which mandated fuel efficiency for U.S. automobiles, demonstrated the effectiveness of energy regulations. Complemented by state-level gasoline taxes, these standards have continued to drive consumers towards more environmentally-friendly, fuel-efficient vehicles.
In the face of our burgeoning climate crisis we can expect similar regulations addressing carbon impacts globally. Recent examples include Denmark’s carbon emissions tax on livestock and California’s Senate Bill 253 (SB 253). The urgency of climate change necessitates robust legislative responses worldwide.
Historically, the 1973 oil crisis favored companies that had already adopted energy-efficient technologies, notably the Japanese auto industry, while the U.S. auto industry, reliant on less efficient vehicles, struggled to recover its industry dominance (Candelo 2019, Kurihara 1984). This underscores the benefits of early adoption of energy efficiency.
California’s SB 253, which requires corporations with revenues over $1 billion to disclose greenhouse gas emissions, is a positive step but could be improved. A broader reporting threshold, similar to the $25 million revenue threshold of the 2018 California Consumer Privacy Act, would be more effective. Greenhouse gases are pollutants, and given the gravity of the climate crisis, we must consider the impact of AI, including from companies with less than $1 billion in revenue.
Smaller technology companies might argue that compliance with SB 253’s reporting requirements is burdensome. However, integrating energy efficiency from the start — like the early adoption seen in Japanese automobiles prior to 1973’s oil crunch — offers competitive advantages. As climate constraints increase, energy-efficient products will be more viable, making early compliance beneficial.
Regulation akin to CAFE standards for AI is likely forthcoming in every jurisdiction worldwide. Start-ups that adopt energy-efficient practices early will be better prepared for future regulations and market demands. Additionally, energy-efficient AI products are more cost-effective to operate, enhancing their appeal to business customers and supporting long-term growth.
Corporate AI policies should prioritize employee education on climate issues to build a knowledgeable workforce capable of advancing sustainability. Product design must incorporate environmental considerations, and operational expenditure (e.g., selecting a cloud service provider) should focus on minimizing ecological impact. Accurate measurement and reporting of environmental metrics are essential for transparency and accountability. Companies should also anticipate future regulatory requirements related to climate impacts and design products to comply proactively. Finally, corporate policy requires avoiding methods that do not yield tangible carbon impact. Adopting these practices will support environmental sustainability and enhance positioning within an evolving regulatory framework.
For AI policies at the individual level, we should all remain cognizant of the environmental impacts associated with artificial intelligence. It’s important to use AI technologies judiciously, recognizing both their potential benefits and their contributions to climate change. Furthermore, sharing this awareness with others can help amplify the understanding of AI’s climate implications, fostering a broader community of informed and responsible technology users. By integrating these personal policies, individuals can contribute to a more sustainable approach to AI utilization.
In Goethe’s poem “The Sorcerer’s Apprentice”, the apprentice’s reckless use of magical powers without sufficient understanding or control leads to chaos and disaster, as “autonomous AI” brooms flood the house with water beyond the apprentice’s ability to manage. This allegory resonates with the contemporary challenges of AI automation. Just as the apprentice’s unchecked use of magic brings unforeseen consequences, so too can the unregulated deployment of AI technologies result in unintended and harmful climate outcomes. Goethe’s poem underscores the necessity of constraining and governing powerful tools to prevent them from spiraling out of control. Effective oversight and regulation are crucial in ensuring that AI, like the sorcerer’s magic, is harnessed responsibly and ethically, preventing the potential for technological advances to exacerbate existing issues or create new ones.
The post Reducing AI’s Climate Impact: Everything You Always Wanted to Know but Were Afraid to Ask appeared first on UC Berkeley Sutardja Center.
