Hacker News new | ask | show | jobs
by vinni2 520 days ago
Perplexity sounds desperate. They got early traction and have money but they seem to be lost.
5 comments

Less desperate, more "there is a point where we needed to stop and we have clearly passed it but let's keep going and see what happens."
I think it is more of, holy shit, OpenAI is entering this space, and maybe Anthropic as well. What can we compete with:

- Name? No

- Technology? No because we rely on other LLMs

Do we have anything? No. We've taken in a lot of money from big-name investors, so let's see if we can turn ourselves into a complementary asset before people realize that LLMs are nowhere near what they are sold as.

Playing with house money for the founders
A few days ago they were talking about making a browser. Smells like it's over for their original business plan. Maybe.
Merge with Arc at that point
Did you miss the Arc memo?
Not OP but what happened?

EDIT: Seems like they abandoned it to... make another browser?

They would be buying one of the largest amounts of user generated data in the world. Sounds good to train on.
The entire point is that Perplexity is several orders of magnitude smaller than ByteDance and just making the offer is a sign of immaturity.
I think Perplexity just wants to stay in the news cycle and get more than it's fair share of discussion.

This is obviously not going to happen, and it's absurd anyone's really engaging with it as a serious proposition. But, because they have a new cycle behind it a lot more people are talking about Perplexity than Claude etc.

Lots of training data for: - Lip syncing - Interpretive dance - Hover text
Electrical/framing/machining/woodworking/physics/tumbling/rockclimbing/instruments/etc/etc/etc.

Honestly it would be a goldmine of training data.

Is there anything of value in TikTok videos?
If you're using the term "value" to refer to monetary value, yes. If you're using it to refer to some other kind of value, it probably isn't relevant to the viability of the proposed merge.
I guess the PP question was like "what real world problems could solve an AI trained on TikTok videos?". Sociology research, maybe.
The training data is an absurd moat for a mega social network, that’s why you save everything forever: the machine learning architecture to exploit it hasn’t even been invented yet and you have longitudinal dissensions all but impossible to get otherwise.

Voice, video, sentiment, ads stuff. With today’s technology you know if a user is about to get cheated on before they do.

The same as an AI trained on YouTube. Extremely valuable.
Electricians showing bench demonstrations of what symptoms faulty neutrals cause, machinists doing step-by-step demonstrations of making tooling, including their fuckups, scrapyards doing deep dives into the in-practice economics of when and why they buy and sell, lawyers giving explanations of the laws involved in popular viral accident videos, baking recipes, comedians showing their material, musicians demonstrating techniques, etc
AI needs user data, user generated content and behavior patterns.

Google has platforms Google also purchased Reddit user content. Meta had platforms and user content.

Well yeah all these AI companies are desperate, it's a highly competitive space with billions of dollars being thrown around to try and win it.