Where is the AI Boom Going?

Summary

This video script details the speaker's travels to the US and Japan, attending the Hot Chips conference and discussing AI advancements. They explore the current state of AI applications, noting mixed success in revenue generation, with companies like ElevenLabs and Character.ai leading in voice and chatbot tech. The script also delves into the AI chip industry, with Nvidia dominating, and speculates on Intel's future. It highlights the significance of the next generation of AI models, like GPT-5, and the industry's focus on scaling and data centers, ending with a reflection on the rapid changes in AI and tech.

Takeaways

• 🌐 The speaker has been traveling to the United States and Japan to engage in conversations and observe the AI industry, particularly around the second anniversary of ChatGPT.
• 💬 The Hot Chips conference at Stanford was a focal point for semiconductor design discussions, with a focus on Meta and Tesla's advancements in ASICs and the potential of Co-Packaged Optics (CPO).
• 🚀 There's a significant interest in AI applications that generate revenue, with a spotlight on Microsoft's OpenAI-powered Copilot products and other startups making millions in revenue.
• 🏢 Companies like ElevenLabs and Character.ai are leading in terms of revenue and usage, respectively, within the AI industry.
• 📉 There's been a dramatic drop in the price of LLM API tokens, with a nine-fold reduction over 17 months, potentially impacting the industry's financial models.
• 🤖 The industry is eagerly awaiting the next generation of Foundation Models, particularly GPT-5, as current models are seen as limited in their capabilities.
• 💰 The financial sustainability of AI companies is a key concern, with a focus on how they can generate returns on the significant investments in infrastructure.
• 🏭 Nvidia's dominance in the AI chip industry is highlighted, with the future potentially hinging on competition from tech giants' AI chips and the success of AI chip startups.
• 📉 Intel's recent financial struggles and the broader implications for the semiconductor industry are discussed, including the potential impact of the CHIPS Act.
• 🚗 Waymo's success in autonomous vehicles and the broader implications for transportation and technology are noted as an underrated story from the trip.

Q & A

What was the primary purpose of the speaker's trip to the United States and Japan?

The primary purpose of the speaker's trip was to attend the Hot Chips semiconductor design conference held at Stanford.

What does Hot Chips conference focus on?

Hot Chips is a conference focused on semiconductor design, featuring a series of presentations in an auditorium that can be attended in person or via video.

What is the speaker's general takeaway from the Hot Chips conference?

The speaker's general takeaway is that companies like Meta and Tesla are making good progress in designing cheap ASICs for specific AI applications, and there is a new optical packaging technology called Co-Packaged Optics (CPO) that might be the future for data transmission between chips.

What is the financial status of Microsoft's OpenAI-powered Copilot products according to the speaker?

The speaker estimates that all the Microsoft Copilots are probably making around a billion dollars annually, which is not considered good enough for Microsoft.

Which company is mentioned as one of the largest in the AI industry in terms of revenue, besides ChatGPT and OpenAI?

ElevenLabs, a voice generation company based in New York City, is mentioned as one of the largest in terms of revenue, possibly making around a hundred million dollars.

What is the speaker's view on the current state of LLM-based applications in terms of profitability?

The speaker views the current state as mixed, with some companies making tens of millions of dollars in revenue, but no clear challenger to ChatGPT in terms of revenue.

What is the significance of the price war in LLM API tokens mentioned in the script?

The price war signifies a significant reduction in API token costs, from about $36 per million tokens in March 2023 to around $4 in September 2024, which could potentially make AI more accessible and affordable.

Why is the next generation of Foundation Models, such as GPT-5, so crucial according to the speaker?

The next generation of Foundation Models is crucial because the current models, like GPT-4, are still limited in their 'intelligence' and capabilities, and the industry is looking towards advancements for further growth and innovation.

What is the current ranking in the AI chip industry according to the speaker?

The current ranking in the AI chip industry is Nvidia first, followed by AMD, with no other significant competitors in terms of AI accelerators available for purchase.

What is the speaker's opinion on the future of Intel, especially after their recent poor financial performance?

The speaker suggests that Intel might continue on its current path, waiting for next-generation process nodes to hit the market, but expresses concern about the financial strain and the pressure from competitors like AMD and Nvidia.

Why does the speaker consider Waymo an underrated tech story?

The speaker considers Waymo an underrated tech story because despite the significant achievement of having self-driving cars on the road, public reaction has been relatively muted, treating them as a common sight rather than a technological marvel.