ShamelessC

This site (ign apparently) is incredibly sketchy without an adblocker on.

Was going to say. Not exactly a hot take.

Why are you using two l’s to spell langchain?

> understand that people should read the full article instead of drawing any conclusion from a mere image, but lots of people don't

People who don’t read the article and find themselves lacking an informed or nuanced understanding of the contents of said article are the only ones to blame here. Full stop.

That HN doesn’t have a culture of shaming those who don’t read the article is partly the fault of the rules here which finds “read the article please” to be a distraction/nuisance but don’t mind that this often results in misinformed discussion propagating toward the top of a thread.

AutoGPT and the like are essentially what OP was referring to. Basically “Siri, but it works and can actually do most of what you ask because the underlying language model is robust rather than a rules engine” (and friends).

There’s caching but only so much can be cached when small changes in the input can lead to an entirely different space of outputs. Furthermore, even with caching LLM inference can take anywhere from 1-15s using GPT4-Turbo via the API. As was mentioned, the more characters you prefix in the context - the longer this takes. Similarly you have a variable length output from model (up to a fixed context length) and so the time it takes to calculate the “answer” can also take awhile. In particular with CoT you are basically forcing the model to use more characters than it otherwise would (in its answer) by asking it to explain itself in a verbose step by step manner.

How is that bible verse relevant exactly?

That’s not an agent based setup.

Is that an agent based setup? Seems like it’s using a few different models wired together manually.

Isn't that just drug substitution?

(hope you're doing well btw)

Certainly what it seems like. They have a model they can't release without backlash, but still want the credit and don't want to wait to receive accolades until they can figure out how to prevent abuse (which I assume is just impossible).

Isn't that true of approximately every voice cloning model released lately? I've seen as few as 3-5 seconds as a matter of fact...

Highly unlikely any non-technical folks ever learn about this, much less try to decipher what it does.

Or perhaps I'm misunderstanding your comment? What do you mean exactly?

It is psychologically addictive as opposed to an addiction which would create a physical dependency in your system. Nicotine for instance doesn't really induce much of a "high" or "euphoria" (many report essentially no effect after prolonged usage). The reason stimulants like adderal can be addictive is that for many they induce an objective euphoria that upon self-conditioning can be incredibly difficult to stop taking the drug even though it doesn't technically have negative effects after stopping for a few days.

Anyone who has suffered from depression or similar and taken an adderal for the first time will understand. It takes many bad feelings and replaces them with an almost manic productivity. There are downsides, but they commonly outweigh the negatives one was experiencing without the drug.

> You should probably not be running your own publicly-accessible ssh servers if this email is not sufficient to at least start figuring out what your next actions are.

That seems like a fairly unreasonable stance.

I believe this was in the late 90’s and in any case, they didn’t succeed at implementing the system they hyped so much - which is the real problem.

Surprisingly hard to find information about this via Google. Seems related to Demis' game company Elixir and their game Republic: The Revolution [0] for which he made some claim that it supported an infinite level of detail/polygons.

Just a cursory glance at the information though. Happy to be corrected or have more detail added.

[0] https://en.wikipedia.org/wiki/Republic:_The_Revolution

> ML on a blockchain

I would actually love it if you had a link with more info on that. Don't take this the wrong way, but my first guess would be that that basically isn't true; either it's not actually machine learning (as is understood today) or it isn't actually a blockchain but rather normal distributed computing being "verified" via blockchain somehow?

Would love to be proven wrong though.

Ah indeed, seems we are in agreement.

With regard to your second point, that seems to be entirely the fault of Dell - no? If they wanted to encourage junior engineers with promotions, maybe they should, you know, promote them? Rather than inventing silly games and trapping them in loops.

I understand your first point however. Can’t imagine how interns and early engineers are supposed to learn without having a highly available set of mentors to learn from.

It is _obviously_ healthy to forget about it and move on. Having said that, any anger at the guy is going to be not because he effected any one person, but because he effected so many - and because he embodied a particular attitude in the cryptocurrency world that lives on in so many others who haven't yet had their luck run out.

The model's weights can be sharded across multiple GPU's. A "common" training server could contain (for instance) eight "A100" GPU's, each with 40 GB (or up to 80 GB) a piece for a total of 320 GB working VRAM. Since they're connected to each other in the same PC, they can communicate with each other quickly enough to calculate in coordination in this fashion. This setup is _very_ expensive of course. Probably in the hundreds of thousands of dollars.

If you're hoping to run the model yourself, you will need enough money and expertise to rent and deploy it to a server with as many GPU's. Alternatively, volunteers and other researchers will be able to quantize (compress) the model and make it easier to run on configurations without as much VRAM.

If you ran it on CPU it may indeed be super slow, but it's possible it's fast enough for the purposes of running the model rather than trying to train that model. I am seeing (limited) success with the maxed out Mac lineup ($4500) using the beefy M1/M2 line of CPU's.

Tactile feedback presumably requires even more powered wearable equipment and will only ever be a lousy approximation of the real thing, short of some sort of brain implant technology.

Quite hyperbolic. Can be summarized as “anti-Apple, anti-VR tech ‘enthusiast’ finds a million reasons to dislike product they never wanted in the first place”.

You probably can't even rent them from Google if you wanted to, in my experience.

Yeah I would have suggested manic episodes (as someone who also has bipolar).

Scaling laws basically guarantee that a sufficiently larger general model will usually beat a smaller specialist model. The misunderstanding is perhaps acceptable but the headline here is essentially restating a well known property of deep learning.

My point was that while it’s not cheating, you can easily infer “what they meant”, and as such don’t need to defend it’s lack of cheating.

I’m pretty sure it goes well beyond ADHD.

That is not a very compelling argument.