[bonoboTP]

I disagree with this. Usually the products are based on published research. This is not easily seen by the enthusiast power user base.

Of course it's only a small fraction of all papers that end up actually being used. Most are mainly about advancing careers and strengthening CVs.

[matusp]

I have been in both academia and industry for years, and I don't think the model you describe is true anymore. It was definitely true 10 years ago, but the situation has flipped. Now, I see really ambitious and impactful research coming out of industry labs. Academia is often lagging behind the state of the art because they lack the resources (data, compute, and skills) to compete.

Academia is also incentivized such that everyone works on the same popular topics to secure grants and citations. This is currently LLMs, where academia needs to compete with multi-billion corporations on a technology that is notoriously expensive. In effect, many researchers work on topics that are pretty non-consequential from the get go (such as N+1th evaluation dataset), but it's the only way for them to stay relevant.

[neilv]

I recently talked with a PI from a well-known university lab, and asked why they were doing a startup, given the ML research problems they were working on.

They said a company was the only way to get access to the compute power they needed for that research.

A startup sounds like probably a good solution, if they get paired with the right product- and business-minded people, and together they find a winning collaboration. (Edit: Or if they get acquired rapidly in the AI boom, and negotiate the right deal to enable their research longer-term.)

[bonoboTP]

A lot of those industry papers are in collab with an academic lab or even often first authored by a PhD student who interns in a big tech lab.

[nis0s]

One key reason you’re wrong is that many interesting things aren’t even getting published, they’re on the DL for years and eventually make it to public spheres and products.

Academia is just a daycare at this point, and many labs shouldn’t exists or get funding. The people who move the field aren’t necessarily the ones with the most citations, they’re usually hard at work in places that don’t publish at all.

[bonoboTP]

Are you talking about just frontier LLM agent stuff or all of the scope of ICML? I wonder what your subfield is.