Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

John Carmack's new AGI company, Keen Technologies, has raised a $20M round

twitter.com

Post reply on HN
Print view View original

191–200 of 643 posts

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#191

Post by keerthiko »

Jack000 wrote: post #116

I'm very optimistic for near-term AGI (10 years or less). Even just a few years ago most in the field would have said that it's an "unknown unknown", we didn't have the theory or the models, there was no path forward and so it was impossible to predict. Now we have a fairly concrete idea of what a potential AGI might look like - an RL agent that uses a large transformer. The issue is that unlike supervised training y…

At least we have lots of very complex simulated or pseudo-simulated environments already — throw your AGI agent into a sandbox mode game of GTA6, or like OpenAI and DeepMind already did, with DOTA2 and StarCraft II (with non-G-AIs). They have a vast almost-analog simulation space to figure out and interact with (including identifying or coming up with a goal).

So while it is significant compute overhead, it at least doesn't have to be development overhead, and can often be CPU bound (headless games) while the AI learning compute can be GPU bound.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#192

Post by jdmoreira »

semi-extrinsic wrote: post #139
Tepix wrote: post #7

I wonder if Carmacks moral compass is in order. First he sticks around at Facebook, now he endangers humanity with AGI. And i'm only half joking.

The meme that AGI, if we ever have it, will somehow endanger humanity is just stupid to me. For one, the previous US president is the perfect illustration that intelligence is neither sufficient nor necessary for gaining power in this world. And we do in fact live in a world where the upper echelons of power mostly interact in the decidedly analog spaces of leadership summits, high-end restaurants, golf courses and c…

Humans can't iterate themselves over generations in short periods of time. An AGI is only bound by whatever computing power it has access to. And if it's smart it can gain access to a lot of computing power (think about a computer worm spreading itself on the internet)

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#193

Post by Isinlor »

KhoomeiK wrote: post #178
Jack000 wrote: post #116

I'm very optimistic for near-term AGI (10 years or less). Even just a few years ago most in the field would have said that it's an "unknown unknown", we didn't have the theory or the models, there was no path forward and so it was impossible to predict. Now we have a fairly concrete idea of what a potential AGI might look like - an RL agent that uses a large transformer. The issue is that unlike supervised training y…

> Now we have a fairly concrete idea of what a potential AGI might look like - an RL agent that uses a large transformer. People have thought Deep RL would lead to AGI since practically the beginning of the deep learning revolution, and likely significantly before. It's the most intuitive approach by a longshot (even depicted in movies as agents receiving positive/negative reinforcement from their environment), but t…

Have you heard about EfficientZero? This is the first algorithm that achieved super-human performance on Atari 100k actions benchmark. EfficientZero's performance is also close to DQN's performance at 200 million frames while we consume 500 times less data.

DQN was published in 2013, EfficientZero in 2021. That's 8 years with 500 times improvement.

So data efficiency was doubling roughly every year for the past 8 years.

Side note: EfficientZero I think still may not be super-human on games like Montezuma's Revenge.

https://arxiv.org/abs/2111.00210

Reinforcement learning has achieved great success in many applications. However, sample efficiency remains a key challenge, with prominent methods requiring millions (or even billions) of environment steps to train. Recently, there has been significant progress in sample efficient image-based RL algorithms; however, consistent human-level performance on the Atari game benchmark remains an elusive goal. We propose a sample efficient model-based visual RL algorithm built on MuZero, which we name EfficientZero. Our method achieves 194.3% mean human performance and 109.0% median performance on the Atari 100k benchmark with only two hours of real-time game experience and outperforms the state SAC in some tasks on the DMControl 100k benchmark. This is the first time an algorithm achieves super-human performance on Atari games with such little data. EfficientZero's performance is also close to DQN's performance at 200 million frames while we consume 500 times less data. EfficientZero's low sample complexity and high performance can bring RL closer to real-world applicability. We implement our algorithm in an easy-to-understand manner and it is available at this https URL. We hope it will accelerate the research of MCTS-based RL algorithms in the wider community.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#194

Post by MichaelCollins »

ryanSrich wrote: post #184
paxys wrote: post #131

Earlier quoted context omitted.

We will have something that we ourselves define to be AGI, sure, but then it's easy to hit any goal that way. Is that machine really intelligent? What does that word even mean? Can it think for itself? Is it sentient? Similar to AI, AGI is going to be a new industry buzzword that you can throw at anything and mean nothing.

> AGI Idk what prompted you to say this, but is there a version of AGI that isn't "real" AGI? I don't know how anyone could fake it. I think marketing departments might say whatever they want, but I don't see any true engineers falling for something masquerading as AGI. If someone builds a machine that can unequivocally learn on it's own, replicate itself, and eventually solve ever more complex problems that humans c…

The way to fake it would be to conceal the details of the AGI as proprietary trade secrets, when the real secret is the human hidden behind the curtain.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#196

Post by roflyear »

jlawson wrote: post #99
mellosouls wrote: post #87

I am not saying the intention here is the same, but the headline doesn't inspire confidence. There's something incredibly creepy and immoral about the rush to create then commercialise sentient beings. Let's not beat about the bush - we are basically talking slavery. Every "ethical" discussion on the matter has been about protecting humans, and none of it about protecting the beings we are in a rush to bring into lif…

You're anthropomorphizing AI and projecting your own values and goals onto it. But, there's nothing about sentience that implies a desire for freedom in a general sense. What an AI wants or feels satisfied by is entirely a function of how it is designed and what its reward function is. Sled dogs love pulling sleds, because they were made to love pulling sleds. It's not slavery to have/let them do so. We can make a ri…

I'm sure cows love to be raised and slaughtered too.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#197

Post by pengaru »

cgrealy wrote: post #72
Tepix wrote: post #7

I wonder if Carmacks moral compass is in order. First he sticks around at Facebook, now he endangers humanity with AGI. And i'm only half joking.

There is a story in “Masters of Doom” about Carmack getting rid of his cat because “she was having a net negative effect on my life”. That’s cold.

It's cold if he killed it/had it euthanized.

Not if he simply found it a better home where it was a better fit and more appreciated.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#198

Post by therouwboat »

dekhn wrote: post #147
cgrealy wrote: post #130

I don’t understand why you would want AGI. Even ignoring Terminator-esque worst case scenarios, AGI means humans are no longer the smartest entities on the planet. The idea that we can control something like that is laughable.

what if humanity's role is to create an intelligence that exceeds it and cannot be controlled? Can humans not desire to be all watched over by machines of loving grace? More seriously, while I don't think it's a moral imperative to develop AGI, I consider it a desirable research goal in the same way we do genetic engineering - to understand more about ourselves, and possibly engineer a future with less human sufferin…

Didn't we have this same talk when Elon thought AI is suddenly going to become smart and kill us all?

Yet my industrial robot at work just gives up if stock material is few millimeters longer than is should be.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#199

Post by viraptor »

stefs wrote: post #172
cgrealy wrote: post #130

I don’t understand why you would want AGI. Even ignoring Terminator-esque worst case scenarios, AGI means humans are no longer the smartest entities on the planet. The idea that we can control something like that is laughable.

the climate crisis might kill us all off if not some deus ex machina (i.e. AGI) comes up with some good solutions fast.

We've already got solutions. We'd only need an Agi to convince people on power to do something about it.

Re: John Carmack's new AGI company, Keen Technologies, has raised a $20M round

#200

Post by sabellito »

semi-extrinsic wrote: post #139
Tepix wrote: post #7

I wonder if Carmacks moral compass is in order. First he sticks around at Facebook, now he endangers humanity with AGI. And i'm only half joking.

The meme that AGI, if we ever have it, will somehow endanger humanity is just stupid to me. For one, the previous US president is the perfect illustration that intelligence is neither sufficient nor necessary for gaining power in this world. And we do in fact live in a world where the upper echelons of power mostly interact in the decidedly analog spaces of leadership summits, high-end restaurants, golf courses and c…

There's a wondeful youtube channel from a researcher who focusses exactly on this topic, I think you should check it out:

https://www.youtube.com/watch?v=ZeecOKBus3Q

Post reply on HN