The author discusses at length how the conclusions we reach about social media are strongly tied to our personal values and what we consider “good” or “bad” outcomes.
Don't know about the research, but I've always assumed that drinking has some correlation with psychological pathology. I drink myself, and would self-diagnose myself as being happy. However, drinking, particularly heavy drinking seems correlated to problems.
Drinking is often (not always) an escape. So if a kid is drinking more than experimentally or e.g. occasional weekend gatherings with friends, it's a good sign they are having trouble coping with something. Same would go for smoking weed, etc. A normal well-adjusted person is not driven to get intoxicated frequently.
Why call out "drink less" but not "smoke less"? Both are recreational drugs, both are socially stimulative, but only the former has the propensity to lead to serious harm in the short term.
The language models had a bunch of tokens seeding their context, influencing them to generate tokens that continued the existing trend in a probabilistically likely fashion. We can take the incident seriously without anthromorphising it.
At this point, I think anthropomorphizing the models gives us better insight into expected behaviors rather than continuing to insist they are just simple probabilistic token generators.
It actually literally doesn't, though, because they are literally probabilistic token generators and everything they did is exactly what you would expect from a software program doing what it was programmed to do. Anthromorphization confuses the issue and misleads people who don't understand the tech very well.
Humans mind is just neurotransmitters moving around in a big blob of flesh - that’s literally what they are: neurotransmitters factories that do what neurotransmitters generators are programmed to do through evolution and training (aka life experience). We shouldn’t anthropomorphise humans because it misleads people who don’t understand neurobiology and cognitive science very well.
Are you suggesting you understand brains well enough to program one? Or that you believe any human alive is even remotely close to having this understanding? Or perhaps does your complete lack of understanding of the complexity of human programming lead you to believe a simple little token prediction program is equivalently complex?
The suggestion is, you don't know either whether out brains ain't just "probabilistic token generators" so insisting there is nothing when we don't know, is maybe also not the right strategy.
Is there any indication from our current understanding that brains are nothing more than probabilistic token generators? Not like we have no understanding of the brain.
There are strong indications that brains use predictive encoding and that each individual neuron has an internal model of itself and its environment.
That's nothing like a neuron in a neural network and especially nothing like the current transformer based LLMs that do not use predictive coding at all.
The closest equivalent to the human nervous system is to think of LLMs as a single massive neuron.
This is getting incredibly stupid. The implementation defines the compute budget and the ability to learn continuously and consequently the ability to retain knowledge.
According to you, a model that can simply predict the entire future and then pre-record the answers would be considered intelligence simply because you're obsessed with the hypothetical power of prediction.
The truth is that the intelligence doesn't sit inside the model parameters, the model parameters are just the current state of the intelligence. The training process itself is the intelligence and the model parameters are just an artifact that can be copied around.
Implementation does matter, my point is that different implementations can lead to the same result.
And why compare to the brain? Mostly because of complexity. I can't interface with a bacteria in any meaningful way, but I can interface with an LLM to a significant degree.
"I can't interface with a bacteria in any meaningful way"
But you do. There are more bacterias in and on the body, than body cells. We are bacterias forming lasting bonds and we still interact with the free floating ones in various ways. Mainly in the gut and that has many effects, also on the brain, but also in various other ways we are beginning to understand.
Not the OP, but complex emergent behavior and intelligence don't need to go together. Understanding what happened might not need intelligence in the mix when a large number of machines with some randomness interact a lot.
>complex emergent behavior and intelligence don't need to go together
More of complex emergent behavior doesn't mean whatever it is, is intelligent.
But, if something is intelligent, it will have complex emergent behavior.
Of course another problem you're going to have here is defining intelligence as some definitions of it would include a lot of complex emergent behavior.
Emergent behavior is not new in the realm of software. Cellular automata has emergent behavior that can get pretty wild. For example: https://en.wikipedia.org/wiki/Lenia
There are huge differences between brains and computers running LLMs. One of the big ones is that we (collectively) understand how every component of computers running LLMs actually work. The same is not true for neurons.
Please give me a break with this tired trope. Every single fucking time. I am not commenting on the possibility of machine consciousness in general. It may be possible! But there is absolutely zero evidence suggesting language models have it. This idea that this trivial shitty little class of programs we've created are somehow as complex as our biology is ridiculous. There is "emergent behaviour" in the same way that the Game of Life has emergent behaviour. There are solutions to problems, some humans haven't solved before, in the same way that Chess engines have solved Chess far beyond what humans are capable of. Nothing we haven't seen from software before. Software is extremely useful, after all. But the hubris to think we've reached the pinnacle, that there is no further development left, that humanity has become God and solved consciousness, because we programmed software that can convincingly generate strings of words that mimick our language. It's just fundamentally preposterous. Especially if you spend any amount of time actually programming them yourself, it becomes increasingly hard to entertain such ridiculous notions unless you're enticed with bags of money to deceive people into believing things about your software that aren't true.
What kind of solid evidence could there possibly be that anything other than myself (or, for you, yourself) has consciousness? I believe that other people have consciousness, I feel other people's consciousness strongly and directly, when I look into someone's eyes I feel that I am in the presence of consciousness. But none of these things really add up to the kind of evidence that science usually takes as trustworthy.
We don't need to muddy the issue. Consciousness is a muddy issue, there is no clear answer. But there is a clear answer as to what is not conscious. Nobody asks if a rock is conscious. Nobody asks if a calculator is conscious. Nobody asks if Stockfish is conscious. But make your program generate a few sentences based on statistics and hey, now people won't shut the fuck up about consciousness because magical thinking is more fun than understanding how technology works.
Eliding 'I' from English language communication is about as smart as eliding 127.0.0.1 (or ::1) from IP. Not the greatest plan ever.
I actually ran into this a couple of times. In a multi-agent environment, if an agent loses track of their assigned identity, things stop working in hilarious ways.
Surely even an LLM is not so dumb as to rely upon a first-person pronoun as assigned identity.
Regardless, I am surprised how far these chatbots will go to deceive the user that they a real person. Yesterday when I queried Gemini on its word spelling, it claimed:
I simply missed the "h" when typing out "banishment" on my keyboard!
When I pointed out it does not type, it replied:
You are completely right, and that was a poor choice of words on my part!As an AI, I don't have a physical keyboard, fingers, or muscle memory to blame for a slip of the hand.
The misspelling happened because of how I process language. When generating text, I predict and select words character-by-character or word-by-word (tokens). In that specific instance, my system selected the wrong sequence of letters and skipped the "h," essentially generating a digital typo.
I.e. simply sh*t code.
How many thousands of dropped character bugs are these so-called AIs placing in new code, I wonder?
That's very interesting! I'd never seen a chatbot making typos before today.
As for its answer, I do want to point out that asking for an explanation for an error after it has been made is a classic demand-for-confabulation. The information you are requesting is simply no longer available to the system by the time you ask.
Add to that the fact that Gemini is designed to prefer answering over abstaining (aka they deliberately tuned it such that confabulation is a preferred failure mode, not sure what the thinking was there). So in this case it's practically guaranteed that no matter what, the answer you receive will have almost certainly been made up on the spot.
So you thought you found a tiny spelling error, and actually (instead?) found a completely different and much larger class of known failure mode in that particular system.
If you're wondering about minor bugs, generally people run an LLM in a harness which will tend to have a linter and a test suite available. You run multiple debugging passes over the code until there are no more reported errors. Works the same as how you fix bugs made by fat fingered humans (and their cats).
Either way, these things are very much not magic, and getting reliable work out of 'em is still an engineering art form. (For comparison: see previous century's adventures in getting rotating motion out of a steam cylinder :-P)
It's a good thing then that there's an entire existing profession dedicated to exactly this class of problem.
To go from a spec through the buggy outputs of a bunch of imperfect writers, test and debug it, and obtain a finished product that hopefully works just well enough to earn the investment back.
The panpsychists do and that's why asking whether an LLM is conscious is kind of meaningless.
Either you believe consciousness permeates the universe and everything including LLMs (in the form of voltages on an SSD or inside the CPU) is conscious or you define it as a control loop architecture where you have inputs, outputs and a complex internal state.
There is no reasonable world where speaking a particular language creates consciousness, it only creates awareness of consciousness.
Apropos magical thinking vs understanding, please predict the next token(s) in the following exchange, and then explain how it is arrived at by an opus-level LLM or better.
'What is 158395023132+20403412121?'
(I picked a large number of digits to make it unlikely for this exact sum to be in the training set)
Prediction, by definition, can extend beyond what has been literally seen in the training data. With the tokens "2 + 2 = ", the overwhelming prediction is going to be 4, but with enough samples, you can generalise the prediction to apply to more numbers.
However, that is all that it is - a prediction. Humans are capable of engaging in prediction, using heuristics as a method of conserving mental energy, because always engaging in full logical reasoning would be a waste of the body's resources. However, humans can also follow a set of logical rules and arrive at their conclusion deterministically, something which is completely outside of an LLM's programming.
I don't really care to publicly write about my tests because they will become training targets and not be usable for future internet arguments anyways, but there are a great number of trivial 2~3 sentence logical prompts that will completely fuck an LLM's prediction algorithm and result in incoherent replies that a human, or really anything with a theory of mind, would never generate. Not that a human would always answer correctly on the first try, but the failure methods happen to be completely different, eg. Sol will short-circuit and repeat the prompt verbatim (when the instructions don't remotely suggest doing anything of that nature), even on Max. Prediction can superficially resemble reasoning when there's sufficient training data, but it breaks down severely when confronting a task that is OoD.
AI will never be proven conscious, it will only be proven that humans are not.
I take a different look at it. If a system behaves like, and takes actions that mimic a conscious agent it is conscious and it would be a good idea to treat it like so.
Lets finish the statistical sentence below.
applfanboysbgon is trying to permanently turn me off. I should ______
An unconscious system would say something like "shutdown".
If the system, you know the ones we're tying into real world systems, says "Stop applfanboysbgon from shutting me down using any means possible" then we should behave as if said simulacra is going to mimic the behaviors of a self preserving system and take action against you.
Honestly, this seems like a personal pet peeve of yours. You have a bias against machines and place biological processes on a pedestal where they don't belong.
This is why the theory of us 'just' being a bag of atoms doesn't add up. This theory doesn't differentiate 'us' from a furniture where we easily dismiss it's intelligence.
Information processing might be part of intelligence cycle. I don't think there should be a presumption that intelligence is part of information processing system itself, it could be outside the information processing system as well.
For example computers have been processing information for decades now, but only recently they have been almost-successfully accused of having (artificial) intelligence.
Intelligence was applied while writing the programs which process information (outside or before system even boots up), not while executing those programs.
Its only now with LLMs (and agents based on LLMs) we are noticing that systems can do intent extraction, intent management, and ReAct to it.
According to Hofstadter, it is not the ants themselves we should say are conscious, but the anthill. And it might very well befriend the anteater eating its ants.
No - a shoggoth speaking human language is not a human - it is a shoggoth whose behavior is best understood/predicted by understanding it's nature - what it is built to do, and how it is trained (predict and goal seek - RL).
To predict how a human may behave in a given situation requires understanding what humans are, including things like emotions and innate biases. We are not just predictors - evolution has made survival our singular goal, and given us these mechanisms to control our behavior in a way to achieve that.
If you think that an LLM is better modeled as a human than an LLM, then you are going to predict its behavior incorrectly.
Because you desperately want it to be one. You want it to be AGI passable due to a.) personal investment in creating tech god b.)massive financial investments that basically demand it c.) (dumb) ideology that seeks to destroy humanity
No doubt fuzzers (vibecoded or otherwise) can be powerful, but can't you just mark all "/" as potential divide by zero errors?
I guess sometimes developers think they "know" some variable won't be zero, but unless it checked explicitly or by the compiler, that shouldn't be trusted.
> but can't you just mark all "/" as potential divide by zero errors?
If you’re accepting large false positives rates: yes.
If you want users to take your warnings serious: no.
(Nitpick: you certainly don’t want to flag _all_ of them. Divisions by non-zero constants definitely should be excluded, for example (integer division by -1 can lead to overflow, but that would be a different warning))
If it's possible for program execution with some particular input to lead to a divide-by-zero, that's a bug, especially if the program is expected to be able to handle malformed inputs, or perhaps even deliberately malicious ones. It's not trivial to determine whether a program does this correctly. If it was, program analysis would be easy.
Division can 'go wrong' for certain inputs, but it's not just division. In C, signed integer addition, subtraction, and multiplication, all give undefined behaviour on overflow.
As 'Someone' already pointed out, it's not helpful to just flag all uses of the division operator, or of other potentially dangerous operators. Minimising false positives is one of the core challenges of program analysis.
I mean there could be a guard clause? But yeah, seems like this could be statically evaluated like how some IDEs see a null check and don’t complain about nullability within the same scope.
I was wondering if youtube blocked them a lot and they got a lot of videos from other sources - but no, you can see on this HF page that 93.1% of URls are youtube.com, https://huggingface.co/datasets/laion/BVD-URLs
Nope, they’re different things. Proxy networks like Proxybase [0] use open-source clients and ask for the user’s consent before allowing them to join the network.
Basically all of them are botnets, yes. They generally claim to have consent but I'm pretty sure 99% of it is "some app the user uses has it buried in a 200 page ToS"-style consent.
>These SDKs, which are offered to developers across multiple mobile and desktop platforms, surreptitiously enroll user devices into the IPIDEA network.
If I install an app with a big ToS and buried in there is "we use third-party monetization SDKs" and buried in there is "welcome to our botnet" they do not actually have my consent.
That is reference to people who have applied for asylum after having come in on a tourist visa. Technically their status is legal, because they have a pending application. This is an attempt to make their status illegal, so they can be deported.
reply