20VC: Bret Taylor: The AI Bubble and What Happens Now | How the Cost of Chips and Models Will Change in AI | Will Companies Build Their Own Software | Why Pre-Training is for Morons | Leaderships Lessons from Mark Zuckerberg
Oct 2, 2024
auto_awesome
Bret Taylor, CEO and Co-Founder of Sierra and former CTO at Facebook, discusses the real state of AI, challenging the notion of an AI bubble. He questions the efficiency of companies building their own software and argues that pre-training models are outdated. As a board member at OpenAI, he claims foundation models are quickly depreciating. Taylor identifies the shift towards AI-driven interfaces and predicts a decline in phone reliance. He also shares insights on leadership in tech, including valuable lessons learned from Mark Zuckerberg.
Bret Taylor argues that investing in pre-training AI models is often inefficient for companies lacking unique capabilities, favoring fine-tuning instead.
Despite skepticism regarding an AI bubble, the economic transformations and potential returns from AI innovations may justify current high valuations.
Organizations are likely to prefer purchasing established AI solutions over custom software development, mirroring trends seen with SaaS adoption.
Deep dives
Skepticism Towards Pre-Training Models
There is a significant skepticism surrounding the practice of pre-training models, especially for companies that are not focused on artificial general intelligence (AGI) research. The belief is that investing in pre-training without unique capabilities leads to inefficient use of capital, much like investing in a custom data center rather than utilizing existing infrastructure. Most software companies should prioritize fine-tuning existing models instead of taking on the additional burden of developing their own. Thus, companies that engage in pre-training without a strong rationale may find themselves regretting that decision as their capital is essentially wasted.
AI in Context of Economic Bubbles
There is an ongoing discussion about whether the current advancements in AI represent a bubble similar to the dot-com era. While some argue that rising valuations and mega funding rounds indicate excess, it is essential to recognize that the economic impact of the internet included many successful companies that originated in that time. Similarly, while present valuations may seem excessive, they might also be justified by the significant transformations AI can bring to various sectors. The key takeaway is that bubbles often have underlying value, and while risks exist, the innovations emerging from AI are likely to yield substantial returns.
Preference for Tailored Software Solutions
Despite the advancements in AI, there is a strong sentiment that companies will prefer buying ready-made solutions rather than developing and managing their own software. Organizations historically tend to regret when they build custom solutions to problems where off-the-shelf software was available. This trend reflects a preference for utilizing established software as a service (SaaS), which maintains lower total ownership costs and operational efficiency. Consequently, as with SaaS's takeover of custom software development in previous years, a similar shift is expected in the AI landscape.
Challenges and Future of Conversational AI
The rise of conversational AI platforms poses both opportunities and challenges for businesses aiming to implement these technologies effectively. Brands will need to navigate the balance between allowing AI flexibility and ensuring they maintain control over how the AI interacts with customers. As organizations like Sierra offer solutions to help brands create their own AI agents, the concept of user agency becomes pivotal; the AI must communicate with empathy while adhering to defined business rules and objectives. Thus, the design of these agents requires careful consideration of how much autonomy to afford them, resembling the tension present in human interactions.
Long-Term Impact and Future of AI Technology
The discussion around AI also touches upon the necessity for robust approaches to navigate future implications of the technology. The potential for AI to reshape customer interactions and streamline business operations underscores the urgency for companies to harness AI responsibly and effectively. As organizations move toward a more conversational interface, it will be critical to establish effective guiding principles for implementing AI solutions. This includes addressing aspects like change management, reliability, and user understanding to ensure that the benefits of AI are maximized and responsibly integrated into everyday business practices.
Bret Taylor is CEO and Co-Founder of Sierra, a conversational AI platform for businesses. Previously, he served as Co-CEO of Salesforce. Prior to Salesforce, Bret founded Quip and was CTO of Facebook. He started his career at Google, where he co-created Google Maps. Bret serves on the board of OpenAI.
In Today's Discussion with Bret Taylor:
1. The Biggest Misconceptions About AI Today:
Does Bret believe we are in an AI bubble or not?
Why does Bret believe it is BS that companies will all use AI to build their own software?
What does no one realise about the cost of compute today in a world of AI?
2. Foundation Models: The Fastest Depreciating Asset in History?
As a board member of OpenAI, does Bret agree that foundation models are the fastest depreciating asset in history?
Will every application be subsumed by foundation models? What will be standalone?
How does Bret think about the price dumping we are seeing in the foundation model landscape?
Does Bret believe we will continue to see small foundation model companies (Character, Adept, Inflection) be acquired by larger incumbents?
3. The Biggest Opportunity in AI Today: The Death of the Phone + Website:
What does Bret believe are the biggest opportunities in the application layer of AI today?
Why does Bret put forward the case that we will continue to see the role of the phone reduce in consumer lives? How does AI make that happen?
What does Bret mean when he says we are moving from a world of software rules to guardrails?
What does AI mean for the future of websites? How does Bret expect consumers to interact with their favourite brands in 10 years?
4. Bret Taylor: Ask Me Anything: Zuck, Leadership, Fundraising:
Bret has worked with Zuck, Tobi @ Shopify, Marc Benioff and more, what are his biggest lessons from each of them on great leadership?
How did Bret come to choose Peter @ Benchmark to lead his first round? What advice does Bret have to other VCs on how to be a great VC?
Bret is on the board of OpenAI, what have been his biggest lessons from OpenAI on what it takes to be a great board member?
Get the Snipd podcast app
Unlock the knowledge in podcasts with the podcast player of the future.
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode
Save any moment
Hear something you like? Tap your headphones to save it with AI-generated key takeaways
Share & Export
Send highlights to Twitter, WhatsApp or export them to Notion, Readwise & more
AI-powered podcast player
Listen to all your favourite podcasts with AI-powered features
Discover highlights
Listen to the best highlights from the podcasts you love and dive into the full episode