If anything, I feel that current breed of multimodal LLMs demonstrate that language is not fundamental - tokens are, or rather their mutual association in high-dimensional latent space. Language as we recognize it, sequences of characters and words, are just a special case. Multimodal models manage to turn audio, video and text into tokens in the same space - they do not route through text when consuming or generating images.
That's the case with all scientific discoveries - pieces of prior work get accumulated, until it eventually becomes obvious[0] how they connect, at which point someone[1] connects the dots, making a discovery... and putting it on the table, for the cycle to repeat anew. This is, in a nutshell, the history of all scientific and technological progress. Accumulation of tiny increments.
--
[0] - To people who happen to have the right background and skill set, and are in the right place.
[1] - Almost always multiple someones, independently, within short time of each other. People usually remember only one or two because, for better or worse, history is much like patent law: first to file wins.
Science often advances by accumulation, and it’s true that multiple people frequently converge on similar ideas once the surrounding toolkit exists. But “it becomes obvious” is doing a lot of work here, and the history around relativity (special and general) is a pretty good demonstration that it often doesn’t become obvious at all, even to very smart people with front-row seats.
Take Michelson in 1894: after doing (and inspiring) the kind of precision work that should have set off alarm bells, he’s still talking like the fundamentals are basically done and progress is just “sixth decimal place” refinement.
"While it is never safe to affirm that the future of Physical Science has no marvels in store even more astonishing than those of the past, it seems probable that most of the grand underlying principles have been firmly established and that further advances are to be sought chiefly in the rigorous application of these principles to all the phenomena which come under our notice. It is here that the science of measurement shows its importance — where quantitative work is more to be desired than qualitative work. An eminent physicist remarked that the future truths of physical science are to be looked for in the sixth place of decimals." - Michelson 1894
The Michelson-Morley experiments weren't obscure, they were famous, discussed widely, and their null result was well-known. Yet for nearly two decades, the greatest physicists of the era proposed increasingly baroque modifications to existing theory rather than question the foundational assumption of absolute time. These weren't failures of data availability or technical skill, they were failures of imagination constrained by what seemed obviously true about the nature of time itself.
Einstein's insight wasn't just "connecting dots" here, it was recognizing that a dot everyone thought was fixed (the absoluteness of simultaneity) could be moved, and that doing so made everything else fall into place.
People scorn the 'Great Man Hypothesis' so much they sometimes swing too much in the other direction. The 'multiple discovery' pattern you cite is real but often overstated. For Special Relativity, Poincaré came close, but didn't make the full conceptual break. Lorentz had the mathematics but retained the aether. The gap between 'almost there' and 'there' can be enormous when it requires abandoning what seems like common sense itself.
It is. If you're at the mountain, on the right trail, and have the right clothing and equipment for the task.
That's why those tiny steps of scientific and technological progress aren't made by just any randos - they're made by people who happen to be at the right place and time, and equipped correctly to be able to take the step.
The important corollary to this is that you can't generally predict this ahead of time. Someone like Einstein was needed to nail down relativity, but standing there few years earlier, you couldn't have predicted it was Einstein who would make a breakthrough, nor what would that be about. Conversely, if Einstein lived 50 years earlier, he wouldn't have come up with relativity, because necessary prerequisites - knowledge, people, environment - weren't there yet.
You are describing hiking in the mountains, which doesn’t generalize to mountaineering and rock-climbing when it gets difficult, and the difficulties this view is abstracting away are real.
Your second and third paragraphs are entirely consistent with the original point I was trying to make, which was not that it took Einstein specifically to come up with relativity, but that it took someone with uncommon skills, as evidenced by the fact that it blindsided even a good many of the people who were qualified to be contenders for being the one to figure it out first. It does not amount to proof, but one does not expect people who are closing in on the solution to be blindsided by it.
I am well aware of the problems with “great man” hagiography, but dismissing individual contributions, which is what the person I was replying to seemed to be doing, is a distortion in its own way.
2024 variant would be, "... do this, you win 1.000.000 points and we pay for your grandma's cancer treatment; fail it, we kill you like we did your predecessor".
2025 gets tricker, as models are explicitly trained to be less gullible and better able to recognize attempts at manipulation, and by today, you'd likely have to be much more clever and probably do a more multi-staged attack - but still, it's always going to be a problem, because the very thing that makes "prompt injection" (aka "social engineering for LLMs") possible is also the thing that makes LLM understand natural language and work as general-purpose tools.
LLMs by their very nature subsume software products (and services). LLM vendors are actually quite restrained - the models are close to being able to destroy the entire software industry (and I believe they will, eventually). However, at the moment, it's much more convenient to let the status quo continue, and just milk the entire industry via paid APIs and subscriptions, rather than compete with it across the board. Not to mention, there are laws that would kick in at this point.
I think the function of a company is to address limitations of a single human by distributing a task across different people and stabilized with some bureaucracy. However, if we can train models past human scales at corporation scale, there might be large efficiency gains when the entire corporation can function literally as a single organism instead of coordinating separate entities. I think the impact of this phase of AI will be really big.
> the models are close to being able to destroy the entire software industry
Are you saying this based on some insider knowledge of models being dramatically more capable internally, yet deliberately nerfed in their commercialized versions? Because I use the publicly available paid SOTA models every day and I certainly do not get the sense that their impact on the software industry is being restrained by deliberate choice but rather as a consequence of the limitations of the technology...
I don't mean the companies are hoarding more powerful models (competition prevents that) - just that the existing models already make it too easy for individuals and companies to build and maintain ad-hoc, problem-specific versions of many commercial software services they now pay for. This is the source of people asking, why haven't AI companies themselves done this to a good chunk of software world. One hypothesis is that they're all gathering data from everyone using LLMs to power their business, in order to do just that. My alternative hypothesis is that they already could start burning through the industry, competing with whole classes of existing products and services, but they purposefully don't, because charging rent from existing players is more profitable than outcompeting them.
Doesn't matter to 99.99% of businesses using social media. Only to the silly ones who decided to use a platform to compete with the platform itself, and to the ones that make a platform their critical dependency without realizing they're making a bet, then being surprised by it not panning out.
They have a clear incentive to do exactly as said - regurgitation is a problem, because it indicates the model failed to learn from the data, and merely memorized it.
> I mean, we all know that they're only doing this to build a training data set
That's not a problem. It leads to better models.
> to put your business out of business and capture all the value for themselves, right?
That's both true and paranoid. Yes, LLMs subsume most of the software industry, and many things downstream of it. There's little anyone can do about it; this is what happens when someone invents a brain on a chip. But no, LLM vendors aren't gunning for your business. They neither care, nor have the capability to perform if they did.
In fact my prediction is that LLM vendors will refrain from cannibalizing distinct businesses for as long as they can - because as long as they just offer API services (broad as they may be), they can charge rent from an increasingly large amount of the software industry. It's a goose that lays golden eggs - makes sense to keep it alive for as long as possible.
Obviously. Those who chose otherwise have all died out long ago, starving to death in their own apartments, afraid that someone might see them if they ever went outside.
Paranoia is justified if it actually serves some purpose. Staying paralyzed and not doing anything because Someone Is Reading Your Data is not serving much of anything. Hint: those Someones have better things to do. LLM vendors really don't care about your bank statements, and if they were ever in a position to look, they'd prefer not to have them, as it just creates legal and reputational risks for them.
> as it just creates legal and reputational risks for them.
Unfortunately I laughed reading this as there is never neither reputation nor legal consequences in the US of A. They can leak your entire life into my console including every account and every password you have and all PII of your entire family and literally nothing would happen… everything is stored somewhere and eventually will be used when “growth” is needed. some meaningless fines will be paid here and there but those bank statements will make their way to myriad of business that would drool to see them
There obviously is reputation and legal consequences. You can get fined for billions for a far more indirect privacy violation that what you are describing. If any big company ever does that, I won't be touching it with a 10 foot pole. And no I don't believe using data for showing me ad is on the same level of privacy violation.
fining facebook 5bn is like fining me $100. and reputation… please… we all know facebook what facebook is/does, they can release secretly recorded phone calls you are making and it’ll be news for like 17 minutes and people will then keep doomscrolling etc
The issue of consequences of data leaks, though real and something I find outrageous, is orthogonal to this discussion. When talking about sending personal or sensitive data to AI companies, people are not worrying about data leaks - they're worrying about AI company doing some kind of Something to it, and Somehow profit off selling their underpants.
(And yes, no one really says what that Something or Somehow may be, or how their underpants play into this.)
people should 1,000,000% be worried about AI company doing something kind of something with it which they are doing as we speak and if not now will be profiting soon-ish
If you think people not using a tool released yesterday are staying paralyzed you must be either working for Anthropic or an enthusiastic follower, in both cases your opinion is not valid. None of this is something that is revolutionary and People have created trillion dollar companies without Claude Max
They somehow have to make big money, so it's just a matter of time until they will sell services to others, based on your personal data. And they probably have some clause in their contracts where you give them the right doing it.
You don't remember when people were generating private keys and tokens using github copilot in the early versions? I'm not sure if they ever completely fixed the issue, but it was a bit scary.
> I am genuinely confused by this comment, given the intensity of disregard/ignorance/bad-faith.
I conversely am confused by the amount of knee-jerk reaction to the word "privacy" people here have.
> I mean we had these before in other very similar topics regarding e.g. Snowden leaks but really a lot of things. So.. uh..
Yes, exactly. Now consider that the world kept on spinning anyway, and the revelations from the aforementioned leaks turned out to have exactly zero impact on the vast majority of people.
To be clear: I'm not questioning the ethical importance of all that privacy talk, just practical importance. It's bad that we don't have more control and protection of our data by default, but at the same time, excepting few people and organizations, the impact is so small in practice that it's not worth the energy spent being so militant about it.
I understand that you have given up and trust me, I can see why one would do that.
That is fine. You can do that.
What is not fine however is discrediting the people that haven't given up as paranoid militant lunatics.
You can be nihilistic, disillusioned, <other adjectives> all you want, but it is not okay to pull other people down and attack them just because they still believe in something you do not appear to be doing (anymore?)
Theoretically, the power drill you're using can spontaneously explode, too. It's very unlikely, but possible - and then it's much more likely you'll hurt yourself or destroy your work if you aren't being careful and didn't set your work environment right.
The key for using AI for sysadmin is the same as with operating a power drill: pay at least minimum attention, and arrange things so in the event of a problem, you can easily recover from the damage.
It’s easy for people to understand that if they point the powerdrill into a wall the failure modes might include drilling through a pipe or a wire, or that the powerdrill should not be used for food preparation or dentistry.
People, in general, have no such physical instincts for how using computer programs can go wrong.
Which is in part why rejection of anthropomorphic metaphors is a mistake this time. Treating LLM agents as gullible but extremely efficient idiot savants on a chip, gives pretty good intuition for the failure modes.
