Are We In An AI Hype Cycle?

Is the latest excitement around AI just another round of dot-com or crypto style hype? The Lightcone hosts discuss where AI might be if the hype cycle is real and what may remain once the buzz wears off.
00:00 - Apply to F24 batch
00:22 - Coming Up
01:05 - Intro
1:21 What the media is saying about AI
2:38 Where are we in the AI hype cycle?
9:32 Where does the value come from?
15:15 - Valuing a Tech Company vs. Speculative Assets
17:51 Comparing the crypto hype cycle to AI now
21:41 Increased ARR
24:14 Why a hype cycle might be good for founders
28:19 Early signs things are working
34:43 The fog of war
37:18 Outro

[chapterme]

Chapters (Powered by ChapterMe) -
00:00 - Apply to first ever Fall Batch
00:22 - Coming Up
01:05 - Intro
01:21 - What the media is saying about AI
02:38 - Where are we in the AI hype cycle?
09:32 - Where does the value come from?
15:15 - Valuing a Tech Company vs. Speculative Assets
17:51 - Comparing the Crypto hype cycle to AI now
21:41 - ARR of the Winter 2024 batch
24:14 - Why a hype cycle might be good for founders
28:19 - Early signs things are working
34:43 - Warren Buffett weighing machine: The fog of war
37:18 - Outro

[Jaacccckkkkkkkkkk]

Consider this: OpenAI was valued at $80 billion in its last funding round, yet there are other models that are nearly on par with OpenAI but valued at under $500 million.

[content-engineering-lab]

AI is probably overvalued at this exact moment in time, but as a founder, I'm not trying to time the market. AI will almost definitely be a much bigger part of the economy in the next 5-10 years. I'm happy to to ride that wave and deal with the ups and downs along the way.

[-BarathKumarS]

LOL, Just look at the YC Batch of 2023 its 90% AI related companies.

Are you guys blind yourselves? Every VC is knees deep into it

[AlexWilkinsonYYC]

I would bet the best returns would be in investing in the companies that provide original data sets to LLM companies rather than the other way around. Would rather be selling shovels.

[MichaelBabich]

The fun fact is that inference is getting cheap really fast and models depreciate in price tremendously. It means that after investing millions and millions in a model one realizes that it lost all its value months after it became public. So, it will be really hard to pay back all those investments. A startup literally needs to increase its audience x1000 from planned to be able to meet the expected number from just a year ago. Big tech uses AI as CAC and marketing tool to attract & retain users. They give away AI for free. It's hard to compete with that for a startup. The only way for a startup is to focus on a really narrow niche with exclusive access to data to be the only player in it. Mass market models by default is a dead end for a startup. Even OpenAI won't hold its position for long. It just doesn't have direct contact with mass audiences like Microsoft or Google. So, people will use Microsoft and Google offerings as default even if they are a little worse than OpenAI's one. The key for a general model is to have access to a mass audience.

[neugey]

If you're asking the question then you already know the answer.

[gregoryneal-qutq]

Gary bro, we are going to apply again we just put the app in pre launch tonight and we want you to actually look at it. Thanks Greg ❤

[AmorosoGombe]

Can't tell you how much I enjoy watching these. Great stuff.

[tedm]

Of course we are, and YC is throwing wood on the fire

[rishanchopra]

Diana's reaction to Garry reciting the cartoon 🤣

[akrebsie]

Yes, but that doesn't mean there isn't a lot of value there. It just means there it is a roller-coaster for a while.

[nokar]

I wonder what the median + percentiles growth looked like for the winter batch. 6 -> 20 million in aggregate is great, but did one or two companies account for most of that? And if so what did they do lol

[Astro_Tutor]

Ai is going to be super helpful for my platform. Definitely looking forward to how my users will interact with it.

[samuelgochhayat]

They knows how AI is superpowerful in coming days

[bbrish]

Garry makes a great point at 9:32
Today, you literally need a computer and an internet connection and you can create an extremely valuable thing with the use of AI

[gptBreeze_io]

Company/products mentioned in the video:

05:47 Nvidia
Became the most valuable company in the world.
Concerns exist around possible overvaluation.

19:01 Coinbase
Not a cryptocurrency itself, but an enabling technology for cryptocurrency.
Has the promise of integrating blockchain into mainstream finance.
Initially viewed as a fringe market but proved to have a clear product-market fit.


33:30 AI-driven Tools for Enterprises
Companies are developing tools for fine-tuning AI using private data.
There’s potential to unlock many applications using AI that haven’t been explored yet.

28:40 PhotoRoom
A company using generative AI for e-commerce image generation.
Valued at around half a billion dollars, demonstrating real revenue generation.

30:00 PermitFlow
Uses AI to automate construction permit processes.
Represents the potential for AI applications in specific industries.

31:34 GitHub Copilot
Noted as one of the fastest-growing products for GitHub.
Accounts for a significant portion of GitHub's revenue growth.

by GPT Breeze

[bwhit]

35:00 “the fast talking Stanford dropout that scams investors out of millions of dollars” LMAO Garry pulling no punches

[abcthegreat]

Build with Cursor and Sonnet 3.5 and then try to call it hype.

[austereengineering]

This happens everywhere. Something steals to much oxygen and a lot great projects are left in the dust.