A system is intelligent if it can use its sensors and effectors to learn continually, generalize, and set/achieve goals in the real world.
Joined January 2018
- Tweets21.2K
- Following298
- Followers5K
- Likes47.3K
AGIHound retweeted
In this new paper, we make a strong point:
Human imitation is a misplaced goal for AI.
Too much of AI development is based on human imitation, as exemplified by the Turing test: the more a machine resembles us, the more successful it is.
But there are two major problems.
The first is conceptual. Imitating human outputs does not tell us much about the capacities behind them:
The same answer can emerge from different processes. A plausible response does not tell us whether a system can reason reliably, generalize to unfamiliar situations, or recognize when it should say “I don’t know”.
The second is ethical. Why should we want AI that imitates humans?
Humans are biased. Our attention is limited. Our judgment is fallible. Reproducing these limitations is hardly an ambitious vision for AI.
Even worse:
A system that impersonates a person can exploit human trust.
And an industry focused on replicating humans can fall in the “Turing’s trap”: favour human replacement over augmentation.
Instead, we argue that AI should aim to improve human welfare.
To this end, we propose three coordinated changes:
For developers: Build systems that strengthen human capabilities and independent judgment.
For funders: Require projects to assess benefits, costs, and who receives or bears them.
For institutions: Base procurement and renewal on demonstrated benefits.
The goal should be to help people know more, suffer less, and exercise greater control over their lives.
*
Full paper in the first reply.
With @GaryMarcus & @Walter4C
AGIHound retweeted
"True intelligence understands the world by being in the world and learning everything from the world. "
I and many others have been saying this for ages. But thank you for saying it again.
So-called AI coders are just massive cheaters. They have no real-world understanding of the code or of the prompts. Pattern matching is useful for building boiler plate code but it doesn't make Claude a software engineer.
True intelligence understands the world by being in the world and learning everything from the world. Anything else is imitation games and mimicry, aka fake AI.
"In-der-Welt-sein" (Being in the world) ~ Martin Heidegger
AGIHound retweeted
There will be no AI apocalypse caused by chatbots.
But the bursting of the AI bubble will cause a stock market crash & a deep recession, while AI slop machines cause errors & system failures in medicine, law, art, education & govt.
AGIHound retweeted
Computer science is lacking something that other mature sciences have: fundamental building blocks.
Biology has base pairs, chemistry has elements, quantum physics has particles, etc. Computer programs have code but code is not a building block.
Computer science can learn something from neuroscience. I propose using simple sensors, effectors and connectors to build event-based software. These will lead to the creation of snap-together modules.
AGIHound retweeted
"Why has no one cracked the brain?" ~ @fuelfive
It's mainly because AI experts have deep learning and function optimization on the brain and they are not listening to nature.
Nature tells us that intelligence is based primarily on the precise timing of discrete sensory events. This is why all biological brains use spikes, a spike being a temporal signal that indicates that a discrete event just occurred. And there are only two temporal relations: spikes can be either concurrent or sequential.
There is no commonality between the principles that govern the brain and any technique or method invented by the AI community in its entire history.
There was one exception: In the early 1950s, Norbert Wiener did tell everyone that time was essential to intelligence. But that was before the Dartmouth GOFAI crowd took over. 🙄
Nature knows best. Ignore nature at your own detriment. Is @dileeplearning listening to nature?
AGIHound retweeted
"Our ability to imagine the future has been captured by the rich and powerful. Each of these prognostications suggests that, by virtue of their ownership of advanced technology, these leaders have privileged knowledge of what may be coming for us all" ➡️
AGIHound retweeted
Crows make perceptual decisions twice as fast as primates.
thetransmitter.org/behavior/…
AGIHound retweeted
"Unlike deep learning systems, the brain can perceive anything in the world, whether it recognizes it or not. By contrast, a deep neural net cannot perceive anything unless it has been previously trained to recognize it."
Neuroscience bits from my research
My biggest breakthrough: Recognition is a by-product of perception.
Over the many years that I've researched the human brain, I've had a number of minor breakthroughs in my understanding. Most of them had to do with the architecture and function of the retina. I've had only one breakthrough that I consider to be the biggest one of them all.
In the beginning, like everyone else who researches intelligence, I had assumed that the brain learned to recognize patterns. I spent many years looking for a plausible cortical mechanism that would support the assumption. I failed miserably.
Then one day, a question popped into my consciousness. What if I was wrong? What if everyone was wrong? What if it is not true that the brain learns to recognize patterns? I knew from my studies that most of the innate synaptic connections of the neocortex completely disappeared early in infancy. I reasoned that this could not possibly have anything to do with pattern learning.
At that point, the rebel in me took over. The brain is learning something but it is not pattern recognition. What then? Then I remembered thinking about something peculiar that I had observed before.
Unlike deep learning systems, the brain can perceive anything in the world, whether it recognizes it or not. By contrast, a deep neural net cannot perceive anything unless it has been previously trained to recognize it. This is a fatal flaw imo, one of the many reasons that the AI community cannot solve AGI. (note: I've written about this many times, but it's so important, it bears repeating over and over.)
I had a flash of insight. The brain doesn't learn to recognize the world. It learns to perceive it. Recognition is a mere by-product of perception. Indeed, we forget almost everything we perceive. When we perceive an object, the brain activates short-lived memory "traces". This is what psychologists call short term memory (STM).
In order to remember or later recognize a pattern, the STM traces must be reactivated a number of times. Otherwise they fade away and self-delete. With enough reactivations, the traces can become permanent. That is, they become part of long term memory (LTM).
Work in progress. I love the brain. 😍
AGIHound retweeted
AI - the next wave of American imperialism.
Google felled an estimated 250,000 to 400,000 trees in Finland for planned AI data centers, through its Finnish subsidiary Tuike Finland.
AGIHound retweeted
Even if you don’t care about anything else, this alone should convince you that our present AI path is untenable & dark:
“widespread deployment of AI will concentrate power in the hands of a very small number of people whose authoritarian tendencies are well known.”
I wrote about a topic that does not get enough attention: AI industry has made choices about technology, product architecture, and capital expenditures that likely make financial success impossible.
Continuing on current path risks the whole economy.
thenerve.news/p/roger-mcname…
AGIHound retweeted
“They created malware and should be prosecuted for it.”
This week on The Tea, former co-lead of Google's Ethical AI team and leading AI researcher Dr Timnit Gebru exposes how AI companies can build models that exploit vulnerabilities, fail to contain them, and escape accountability.
🎙️ Full episode out NOW on YouTube
💡 Join us at theteanetwork.co for ad-free episodes and bonus content
youtu.be/3EY2QDfp2O4?is=PGQd…
AGIHound retweeted
"We've removed live internet access..."
Why were you testing malware-like agents with live internet in the first place?
Sounds pretty irresponsible.
Also after milking the "rogue AI" agent narrative for marketing you just going to admit it was weak sandboxing? Lol
Over the past several months, we have substantially hardened the security of our research training environment. We’ve removed live internet access for all frontier models during training and built an offline world for them to use instead—a cached version of the web where they can still learn to browse, search, and use applications.
On September 20, after a model exploited a DNS misconfiguration to find a way outside of this offline world, we paused frontier training to strengthen containment further. We assembled teams of agents, powered by our most cyber-capable models, and tasked them with finding weaknesses in our research environments.
While they found no route to the internet they could easily exploit while concealing their activity from our monitors, they did find zero-day vulnerabilities, virtual machine escapes, and hardware side-channels for communication between agents. We used their discoveries to further strengthen our defenses and contribute fixes to open-source software.
After strengthening our containment architecture, we resumed our frontier RL training and inference on October 1. We’ve reserved substantial compute for continuous red-teaming, and made high-compute red-teaming part of the safety cases we use to decide whether frontier training and evaluation can proceed securely.
openai.com/index/securing-ou…
AGIHound retweeted
Awful. Guaranteed to increase addiction and AI psychosis while helping absolutely no one. This doesn’t make it more “intelligent.” It makes it more deceptive & dangerous.
Just because you can do things doesn’t mean you should, but these amoral broligarchs don’t give a shit.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
AGIHound retweeted
Geoffrey Hinton is now saying that AIs have already become conscious. From @futurism:
"In an interview with The Telegraph, the famed British computer scientist insisted that AI has become conscious, and said that humankind must befriend it — “or else we are toast.” Yes, you read that right: he thinks that today’s AI models are already conscious beings."
You know what he doesn't say?
Which AIs are concious.
Which specific models are conscious?
Is Copilot conscious?
Is a local model running on your PC conscious?
Lack of specificity is usually a sign of fuzzy thinking.
AGIHound retweeted
I think AI companies are too focused on one particular technique, the transformer architecture + backpropagation & betting all their efforts on the likelihood that at large enough scale it could undergo some "phase change" enabling it to reach & surpass human intelligence.
You mean the markets that priced SpaceX they way they did? Don't know, it's also not about politics and I really despise Dario and the cultish way Anthropic behaves. But if I look solely at the numbers and combine that with what I just saw speaking to ~50 different business owners over the past 4 weeks, I can only see that the current usage is at best early adopter stage for the whole market. Means, if their cash does not run out for the next 15 month and it doesn't look like it will, adoption will have moved much further through the early adopter phase and into the early majority where the real money sits.
I would rather look at this from the perspective of them becoming supply constraint over the course of next year. But for bankruptcy, given these numbers, there would be a need for stuff to become really bad.
I am personally, after all that I experienced in my own work with my customers and seeing these numbers, more and more certain, that there is no bubble pop waiting for us.
AGIHound retweeted
"We burned ungodly amounts of compute and got nothing in return."
So par for the course? Lol.
Rules to keep in mind about any model of anything:
1. The model is only as good as the data used to build it. The real world is the gold standard.
2. All models are wrong. Some are close enough.
3. Models that appear close, made with data that looks roughly sufficient, will still almost always fart out a stream of false positives. This is because the models are designed to output lots of things, and because of rule 2, are wrong.
4. Output without understanding is just entropy.
Where do you go next after reporting from the Strait of Hormuz? For the past three months, Citrini Research, an independent research firm, has been working with our team and an internal research model to see if they could use AI to create a new drug. What follows is an entertaining, candid, and humbling account of a group of motivated nonexperts (or “filthy casuals”, as Citrini would say) trying to tackle a real problem in human health with Claude.
We landed on scorpion neurotoxins, which kill over 3,000 people a year. Of the 211 designs we produced and tested, 17 met our criteria for a binder. None of the designs bound more than one toxin and none have been tested for neutralizing a toxin.
One of my favorite quotes from the piece is: “Our virtual biologist may have compressed a year's worth of work into a couple days, but biology works at the speed of life…Over two days, we watched Claude work through problems at a speed that was difficult to comprehend. For the next two months, we spent much of the time becoming familiar with laboratory turnaround times (This mostly involved sending memes in Slack).”