- Lore
- Posts
- What Is Going On At OpenAI?
What Is Going On At OpenAI?
And Why Logan Kilpatrick Left OpenAI for Google...
Hi everyone š,
Iām trying a dramatically different newsletter format this week. Please let me know what you think!
I want to get back to having fun with the newsletter. Sharing my real, unfiltered thoughts on the most critical things in AI and tech.
With my experience creating VC-backed startups in Silicon Valley, surviving a tornado and being trapped under rubble as a child, making money as a gamer as a teenager in Alabama, and now living in Kyoto and doing a podcast with Matt Wolfe and HubSpot, I have some unique insights and points of view to share.
I also want to focus more on supporting the techno-optimist and e/acc communities and helping others, not just sharing the same AI news everyone else is.
Every week, I'll share the 5 most interesting or important things I find, along with my thoughts on what you need to know. This could include technology breakthroughs, Silicon Valley rumors, anti-AI regulations, and how we can push back against them. And how I'm using AI to make money and get jacked in Kyoto.
If you're looking for a comprehensive overview of all of the latest AI news, I suggest checking out these newsletters from my friends:
Future Tools ā Twice a week newsletter with the best list of new AI tools and AI news from my podcast co-host Matt Wolfe.
The Rundown ā The most comprehensive AI daily newsletter.
The Neuron ā Best combination of AI news and insights. Plus, theyāve got a kawaii cat logo.
Please reply to this email or tweet me your thoughts about this new format! I will keep the intros super short in the future ā and spend a lot more time on editing. ;)
This issue is brought to you by Graydient, created by my old game-industry friend Niero.
Graydient is your ultimate destination for unlimited Stable Diffusion and unfiltered language models on all your devices. Sounds pretty cool, right?
Sign up now to get 2 months free, and save 15% off your subscription.
What Is Going On At OpenAI?
A conversation with OpenAI on what to expect in the next 12 months
- today's systems are laughably bad
- ChatGPT not long term engagement model
- models capable of complex "work"
- like a great team mate working with u
- shift towards verbal interfaces & beyond, Multimodalityā Źį“É¢ÉŖį“ (@legit_rumors)
3:11 AM ā¢ May 7, 2024
Another day with a ton of hype coming out of OpenAI, this time at the Milken Institute. I hope the hype is true, but I am starting to wonder what the f**k is going on over there.
Iāve been in the āGPT-5 is just around the corner and is going to blow us awayā camp, but after my interview with Logan Kilpatrick (details below), Iām not so sure.
Hereās a brief overview of the last month:
Sam Altman has had multiple interviews where heās said GPT-4 basically sucks, and GPT-5 will be way better. My favorite recent one was the 20VC interview.
OpenAI was supposed to release a search product this week, as reported by Pete at The Neuron, but it might be delayed.
GPT2, a mysterious LLM, was released on HuggingFace without any explanation of who created it or what it is. In some areas, it seems to be at the GPT-4 level or even below. However, in other areas, people have had some amazing results.
Sam Altman has tweeted several cryptic tweets referencing GPT2. Then, multiple different versions of GPT2 showed up, all seeming to have their own strengths and weaknesses.
My best guess about GPT2 is that itās a fundamentally more efficient model architecture that OpenAI created, which probably is the basis for how they built GPT-5.
If thatās the case, tho, itās still really odd how theyāre releasing this in such a mysterious way and why theyāre doing it since itās clearly not GPT-5. Unless theyāve WAY overhyped GPT-5, which I still donāt think is the case.
I'm unsure if this is some hype marketing strategy dreamed up by GPT-5, a way for Sam Altman to buy time or just a result of me spending too much time on Twitter.
I still feel like OpenAI has something incredible coming with GPT-5. But, I am a bit less confident than I was a month ago about that prediction after talking with Logan Kilpatrickā¦
I am glad, tho, that it appears that Sam Altman is finally putting out a more techno-optimist vibe.
using technology to create abundance--intelligence, energy, longevity, whatever--will not solve all problems and will not magically make everyone happy.
but it is an unequivocally great thing to do, and expands our option space.
to me, it feels like a moral imperative.
ā Sam Altman (@sama)
7:59 PM ā¢ May 4, 2024
I hope heās got something amazing cooking.
Why Logan Kilpatrick Left OpenAI for Google
In this weekās episode of The Next Wave, Matt Wolfe and I spoke with Logan Kilpatrick before anyone else about why he left OpenAI for Google. If youāre unfamiliar with Logan, heās a huge advocate of open-source software with an incredible background at companies like NASA. And heās the President of NumFOCUS, a non-profit supporting open-source in data, research and science.
But to me, for the last year, Iāve always just known him as the main person at OpenAI that was super friendly and easy to reach and talk to on Twitter. For many people, he was the human voice of OpenAI, so itās a big deal that he left.
In the video, we cover a ton of other stuff, including open-source vs. closed-source, how ChatGPTās real origin story isnāt what youāve heard, and more!
I hope youāll check it out. Itās blowing up with over 4,000 plays in less than 24 hours!
P.S. It would help if you subscribed to The Next Wave on YouTube, Apple, and Spotify. Thatād make me, and HubSpot, happy. And if you like, leave a comment or review. Unless itās negative, then DM that to me. š
User Experience More Important Than Ever
The rise of AI means *great* software is necessary again. After spending all day looking at demos so far today, I can say resolutely:
Founders must learn how to make good user experiences again.
I can tell immediately. I only want to fund people who can make great software.
ā Garry Tan (@garrytan)
1:10 AM ā¢ May 5, 2024
I agree strongly with this tweet from Garry Tan, president of Y Combinator.
Today, I watched this YouTube episode from Matt Wolfe, where he tried many AI tools. And damn, some of them were incredibly poorly designed.
Reminds me of a conversation we had with Greg Isenberg recently, where we talked about in the Age of AI, when everyone will be able just to copy your company or SaaS tool with the use of AI, you need a competitive edge.
In the episode, we talked a lot about community and social media followings being new kinds of moats. But I think we should have talked more about user experience, too.
I donāt want to put down any specific companies that have bad products, but here are some examples of what nailing user experience with AI looks like:
Examples
Magnific by Javi Lopez, which was just acquired by Freepik. An incredibly beautiful, simple tool for upscaling. Where most saw the lack of consistency in AI art and upscaling a downside, he turned it into an upside by presenting it as the AI being creative and magically improving your art or photos.
AI Dating coach built on-top of ChatGPT making $190,000 a month. This kind of product couldnāt have existed without AI.
Midjourney. The #1 AI art tool, created by an insanely small team(about 40 people now) and will soon if not already be worth over a billion dollars. Oh, by the way, if you havenāt tried Midjourney in a while, they finally have the web version working, so you can use it without Discord. Itās MUCH better. Check it out.
As this list shows, if you nail user experience and leverage AI, you can build an insanely profitable company with few people.
Itās worth thinking deeply about before you just tack AI on to a product, put a āØ on it, and call it a day.
How can you make a new user experience that wasnāt possible before?
For example, if I were Airbnb, I would not just think about how to improve the Airbnb website with new AI features. Iād be thinking about the entirely new experience of finding a place to stay now that AI exists.
I think most companies, instead of tacking on AI, should be building internal skunkworks that test out AI and build entirely new products, even if those products might kill their own company. Because if they donāt do it, a small team of motivated people powered by new AI tools will be coming to do it.
But, in the classic innovatorās dilemma style, most companies āknow thisā and still wonāt do it.
The State of Innovation At Apple
This week, Iāve heard many rumors about Apple making big moves in AI. The biggest rumor is that Apple might build its own Silicon-based AI servers instead of using NVIDIA. While this is a step in the right direction, I believe Apple needs to be taking AI more seriously.
I mean, they recently stopped development of their decade long car project, which was supposed to be one of their big products to take on Tesla. And a project that would have given them massive amounts of real-world training data. And, then more recently it was announced that theyād just partner with Google or OpenAI for their AI instead of doing their own.
This tweet from YouTube co-founder Chad Hurley perfectly encapsulates how I feel about Apple right now, and the humor of it all.
The big boys mean business, Apple heard about all those useless AI pins trying to take their crown. Oh yeah? Hereās a thin iPad, our thinnest product yet! Innovation isnāt dead. Ok, Siri has been a dumpster fire but weāre a monopoly, weāll just get Google or OpenAI to fix that.
ā Chad Hurley (@Chad_Hurley)
4:37 AM ā¢ May 8, 2024
OpenAI, Meta and others out here about to democratize access to intelligence and make entirely new forms of computer interaction possible. And, Apple is still playing the whoās got the thinnest StarTrek screen game like thatās exciting to anyone now.
I know because Iām from Alabama, and Tim Cook, the CEO of Apple, went to Auburn, Iām supposed to be rooting for him. And, he seems like a great guy and an excellent operator.
But I still believe that his being the CEO is a huge mistake. Apple seems to be out of good new ideas, and no one there knows how to go from 0 to 1.
What was the last good new product they actually released? Apple TV failed. The Apple Vision Pro is cool tech, but itās an expensive niche product at this point. Steve Jobs would have never released it without a killer first app and at a price point that people could barely afford, but still was within reach.
Itās like Boeing having all these issues with their planes and their stock tanking because it has a CFO as the CEO. And the CFO thinks planes are just numbers in a spreadsheet. Apple has the same problem. Theyāre running it like a giant spreadsheet business instead of a product incubator. And in the age of AI, when everything changes fast, the spreadsheet companies will lose.
I hope Apple can correct course and have another great Apple rebirth. But the last Apple rebirth required Steve Jobs, and unfortunately, heās no longer with us.
Things Iām Reading Or Watching
7 Fascinating AI Tools You've NEVER Seen Before (Underground AI #4) - Love this new format from Matt where he's trying out indie AI tools and being very critical of the ones that just arenāt necessary.
America Needs More Techno-Optimism, with Marc Andreessen and Tyler Cowen - Great interview with Marc Andreessen about why techno-optimism is critical for our future as a species.
The King of Internet Writing | Paul Graham | How I Write Podcast - I know the quality of my writing needs to improve a lot now that I'm taking back over writing the newsletter. So, Iām learning from some of the greats by listening to Dave Parrell's podcast. While listening to his pod, I was pleasantly surprised to hear my friend Megs mentioned and her AI art style techno-elysian, which she has allowed us to use as part of the aesthetic of Lore.
Thatās all for today.
Letās keep accelerating and warding off the decels together!
-Nathan
Find me on X, LinkedIn and Instagram
Watch The Next Wave on Apple, Spotify or YouTube.