Not sure why this is linked. The field has kind of moved on from DAPO

where is the bleeding edge now then ???

Super cool work! What's next? Do you think with longer memory limits more interesting programs might emerge? Or is it substrate dependent?

there are many bottlenecks in the substrate - memory limits being one of them. other things like the A-B concatenation inherited from the bff paper also severely limit the possible dynamics. understanding and avoiding these bottlenecks are some of our next steps!

Your website is broken btw - Clicking into another article through related work goes to the same page

Pretty cool! Simulating artificial cells and observing their complex emergent dynamics has been an interest of mine On a similar vein, have you seen this before? https://github.com/chrxh/alien

If you mean this in a psychedelic context, my theory is the temporary dissolution of this “society” in recognising the “self” as “self”

Considering that brainfuck only has 8 characters and models are scoring at 6.2% I don't think tokenization is the issue

The only issue. *

Reasoning is hard, reasoning about colors while wearing glasses that obfuscate the real colors... even harder... but not the core issue if your brain not wired correctly to reason.

I suspect the way out of this is to separate knowledge from reason: to train reasoning with zero knowledge and zero language... and then to train language on top of a pre-trained-for-reasoning model.

LLMs already use mixture of experts models, if you ensure the neurons are all glued together then (i think) you train language and reason simultaneously

I think the product was just too early for its time, and there is not much demand for it. For what it's worth, the founder (Ren Ng) went back to academia, and was highly instrumental in computer vision research, e.g. being the PI on the paper for NeRF: (https://dl.acm.org/doi/abs/10.1145/3503250)

I don't think it was quite too early, it just makes tradeoffs that are undesirable.

Lytro as I understand it, trades a huge amount of resolution for the focusing capability. Some ridiculous amount, like the user gets to see just 1/8th of the pixels on the sensor.

In a way, I'd say rather than too early it was too late. Because autofocus was already quite good and getting better. You don't need to sacrifice all that resolution when you can just have good AF to start with. Refocusing in post is a very rare need if you got the focus right initially.

And time has only made that even worse. Modern autofocus is darn near magic, and people love their high resolution photos.

I find it very useful for wildlife photos. Autofocus never seems to work well for me on e.g. birds in flight.

It's also possible to generate a depth map from a single shot, to use as a starting point for a 3D model.

They're pretty neat cameras. The relatively low output resolution is the main downside. They would also have greatly benefited from consulting with more photographers on the UI of the hardware and software. There's way too much dependency on using the touchscreen instead of dedicated physical controls.

> I find it very useful for wildlife photos. Autofocus never seems to work well for me on e.g. birds in flight.

The more recent cameras can detect birds specifically and are great at tracking them.

> It's also possible to generate a depth map from a single shot, to use as a starting point for a 3D model.

That is true, but is a very niche need. Wonderful if you do need it, but it's a small market.

There is a limit to the resolution needed by consumers, so in that sense maybe they were too early.

I'd argue the opposite, consumers need more resolution than pros.

A pro will show up with a 300mm f/2.8, a tripod, a camera with good AF and high ISO, and the skills, plan and patience to catch birds in flight.

But all that stuff is expensive. The consumer way to approximate the lack of a good lens is a small, high res sensor. That only works in bright light, but you can get good results with affordable equipment in the right conditions. Greatly reducing the resolution is far from optimal when you can't have a big fancy lens to compensate.

And where is focus the hardest? Mostly where you want to have high detail. Wildlife, macro, sports.

Was going to comment the same but you beat me to it!

On that note, reading the ChatGPT-esque summary in the linked article gave me more brain damage than any AI I've used so far

The irony. It isn't even a new study. Way too much has been written about this flawed study when we should just be doing more studies.

I am sorry to hear that OP, I hope you fight the good fight and wish you all the best.

On a different note, a quick cursory glance of this company really makes me wonder who even gave them $160M? The company site is soulless and filled with corporate jargon, and the whole company smells of bloat and leadership team is a long list of people in bullshit jobs. Is this where VC money goes these days? I am dumbfounded by the degree of mismatch between capital and utility

Well, a cursory glance into the funding round shows an equity firm (Highland Europe) which had one of their partners moved into a director position at Deepski. Could be the guy collecting "AI leadership experience" for his resume.

Another notable investor is a french public entity (bpifrance) which might very well have similar reasons but on the country level, having to allocate funds to "AI" to demonstrate France leading role in future technology.

Note that this doesn't mean Deepski and it's leadership can't be great - but the thought experiment of some well networked people noticing they could all benefit over a glass of wine also doesn't seem too far off.

Edit: Maybe there's an angle for someone really serious about this FOSS dilemma here, I hear public entities really hate bad PR - maybe ask bpifrance how they feel about this?

If they do some decarbonization real estate SaaS and need connections with both regulators or whoever issues certificates and REITs etc, then it makes sense. I don't think they need a lot of software for that.

> I am dumbfounded by the degree of mismatch between capital and utility

That's often a sign of money laundering