Hacker News Reader

Comment by preommr | Hacker News Reader

This is missing the most important requirement: Be an AI company.

I think it's likely everyone is overindexing on this, in every direction. The modal reason any startup applies to YC is to line up fundraising. Investors are obsessed with AI right now. Yes, they shouldn't be. At the same time: the CS techniques underneath "AI" are the most important thing to happen to CS in decades; it would be a little weird not to somehow intersect "AI", if only because you're working with embeddings or high-dimensional data somewhere; it would be like building an application purposefully without topological sort, or without trees. And if you're doing anything like that, it would be pretty dumb not to play up the "AI" angle in this market.

I'm sure a lot of it is pure slop, just a bunch of prompts and a CRUD app. But I wouldn't be so quick to assume that anybody talking about "AI" is just a bunch of prompts.

I thought you were just being snarky, but I went and looked, and wow, it really is all "AI": https://www.ycombinator.com/companies?batch=F24

Without looking carefully, I can't tell which ones are generative AI / boiling the oceans to bullshit you; it's possible that some of these are legit uses of machine learning, which existed before ChatGPT and will hopefully continue to exist in the future. But I bet it's all ethically questionable bullshit.

Forerunner AI jumps out as a likely ethical black hole, based on its one line description "Copilot for aerospace engineers making rockets, munitions, satellites." On top of code theft, water use, emissions, and the many other horrors of Copilot and its ilk, we can also add war profiteering.

EDIT: Prediction: YC's next batch will include a startup trying to replicate whatever the fuck this is: https://www.972mag.com/lavender-ai-israeli-army-gaza/

One of the most impressive companies in my batch (W24) had nothing to do with AI, or indeed software: https://www.astromecha.co/

So no, not a requirement. OTOH if you do come with an AI idea, they won't reject you just because they already accepted a few hundred others :)

maybe that's why it was impressive/memorable?

No, I do not think so. Tom accepted us 80% because of our team, not the idea.

Honestly this is the same level of snarkiness of saying, "Be a web company." in the early 2000s

YCombinator is a tech investor, of course they are investing in the future. It just so happens that in this current tech cycle AI is the future.

> YCombinator is a tech investor, of course they are investing in the future. It just so happens that in this current tech cycle AI is the future.

I think the term you're looking isn't "future," it's "hype."

I am sure wheelwright's felt the same way...

I just saw a realtime generative minecraft demo, your head is in the sand if you think this is hype.

> I just saw a realtime generative minecraft demo

Wait, that demo wasn't a joke?

I feel like that is a great example of hype. Some that looks impressive but once you watch for a bit you realize it's a shallow copy that required a million hours of footage to generate.

So much pessimism and downright denial on the impact of this technology.

The old guard have put earplugs in and wonder why everyone's dancing, they just can't hear the music. In this context it's their refusal to try anything and attempt to downplay everything. The technology will move on, with or without them, and that's the most beautiful part.

You can hate the marketing of AI, you can hate people starting startups to take advantage of this new technology, but you can not stop this technology from moving forward. They will fall into the same category as the people who denied the impact of the Microcomputer, the Graphical User Interface, and the Smartphone.

Personally, I didn’t find the Minecraft demo very impressive, aside from the speed of inference. 20fps is quite an achievement, generating Minecraft screenshots, not as much.

No. I think we're past the point of it being hipsterism to suggest that AI techniques are "hype". Not every startup exploiting it is bona fide, but what's happening here is for real.

I second this. I just fear the hype surrounding generative AI is leading to extreme tunnel vision and neglect of other promising tech (cough crypto) and even other areas in ML and deep learning that might actually be more fruitful for Agi.

(If it helps anyone else, I'm completely dismissive of the value of crypto).

The best thing to come out of crypto hype is the attention to Markel trees.

That probably played a part I. The development of CRDTs, which are straight up magic.

The ratio of hype to substance is ridiculously out of this world. But like GP said (about AI), there's something there underneath all the hype

I'm not sure the history of CRDTs is at all bound up in cryptocurrency, which, for instance, isn't mentioned in the INRIA paper (which paper mentions previous CRDT designs that just didn't have the name).

Yes it's a future that's not happening and would soon be past.

Reply on news.ycombinator.com