OpenAI plans to ring in the New Year with a for-profit push Amid growing competition and skyrocketing compute requirements necessary to support the next generation of AI models, OpenAI is shaking up its corporate structure - again. The ChatGPT creator on Thursday outlined its plan to establish a public benefit corporation (PBC), which it argued would clear the way for large-scale …
The article uses this picture. Isn't that a son of Martin Gore? That face looks soooo alike, that cannot be an accident...
I think any investors they have will be fine with this*. It lets them get lots more money, and they need lots more money so that Sam Altman can get lots of it, sorry I mean they can quickly build the most amazing models possible before someone beats them to it. That's necessary if OpenAI is ever going to be profitable, and if you thought they weren't going to, you wouldn't be investing. At this stage, investors are still banking heavily on outspending and outlasting their competitors, so the need for more infusions is clear. Any investor who thinks they've stagnated is not too interested in still being an investor either, and the more for-profit they are, the easier it would be to find someone dumber to take the stake off you.
* All investors except, of course, for Elon Musk if you can call him one now, but that's not really about a for-profit move even though it's probably the lawsuit he's got the best argument for.
It's not really Open. It's not really 'artificial intelligence', just copy and pasting with much thesaurus use. And there isn't much public benefit.
I think MS are looking at the raw cost of using a supercomputer to fix an overly long paragraph and have decided that half a dozen lines of code would do that for free. Which is sensible. Crap, I used the words Microsoft and sensible in the same sentence. Shame on me.
You really don't need to dim the lights somewhere just to correct grammar. Ditto if you wanted to know when they first made Angel Delight. Just go to Wikipedia. At a time when we are told that just using a cotton bud kills a rainforest pixie, running hundreds of thousands of GPUs to give the semblance of a human typing a response to someone too lazy to type in a suitable search request is just barking. If the penny drops out there on this, those billions will have been spent on e-waste. If they haven't already been. How long before a GPU is obsolete and your data centre is an ace caff with quite a nice chip museum attached? 4 years?
The insane cost of running their LLMs (just the energy cost per task are insane, add to that the cost of the hardware and cooling) coupled with the mediocre output (Futurism: OpenAI Research Finds That Even Its Best Models Give Wrong Answers a Wild Proportion of the Time, Ars Technica: Apple study exposes deep cracks in LLMs’ “reasoning” capabilities Wired: Google, Microsoft, and Perplexity Are Promoting Scientific Racism in Search Results) means that I doubt they could reduce the cost enough, or charge users the required thousands a month per seat, to ever break even.
LLMs appear to be a dead end on the road to artificial intelligence and eventually investors will start to notice that too, some have already begun to doubt the "AI" hype.
"They're expecting people to PAY for this dross?"
Indeed they are, up to $2k per month according to some recent reports. Nice work if you can get it!
In fairness, ChatGPT and its ilk have their uses. The trick is to use them as a time saver to handle tasks you know you could do yourself, and are therefore in a position to catch any silly errors. Case in point, I used it to rework a few hundred lines of Python recently. Yes, I certainly could have done it myself - I wrote the original, after all - but it would have taken me an hour or more, whereas it took me a couple of minutes to paste the original into ChatGPT and tell it what I wanted to change. The output was exactly what I asked for, and ran correctly first time - possibly because I specified exactly what I wanted.
Similarly, I had a line in a batch file that just wasn't working and, not having written a batch file for many a year, I couldn't see why. I could have posted a question on a forum, then waited for some kind soul to point out my error. Instead, I asked ChatGPT and got an immediate - and correct - answer.
Used like that, they're a handy tool in the box. But letting them run amok, doing god knows what with little or no supervision, no. That way madness lies. They certainly do make mistakes, and shouldn't be trusted with anything you can't do yourself!
Agree with the main thrust of your argument BUT the issue I have is the 'supervision' requirement !!!
You just know that some 'eager bod' is going to push the boundaries too far and let the requirement to supervise the output go unheeded !!!
This is a given and is the reason I cannot appreciate so called AI as it is currently manifest, it 'will be/is' misused/abused with no regard for the consequences.
The pitch given regarding what AI can do is nothing more than a scam, there is no known solution to the 'Hallucination' problem.
Being 'right' some of the time is not good enough as some people will not be able to see the 'wrong' answer and will treat it as gospel.
AI is still a Sci-Fi future yet to happen !!!
:)
The pitch given regarding what AI can do is nothing more than a scam, there is no known solution to the 'Hallucination' problem.Being 'right' some of the time is not good enough as some people will not be able to see the 'wrong' answer and will treat it as gospel.
AI is still a Sci-Fi future yet to happen !!! ..... Anonymous Coward
Bravo, AC, and take a bow, for that was quite close to honestly being the gospel truth driver which reveals and delivers Presents and Easter eggs for AI is a SCI-Fi Future happening......... and way beyond any possible conventional and traditional remote Earthly human command and control instruction, direction and production interference too which you might like to realise and accept is something of an Almighty Heavenly Gift given the inherent insanity currently prevalent and feeding and presiding in the barbarous nature of the native inhabitants in residence.
This post has been deleted by its author
It's the requirement for "supervision" and "careful checking of the output" that will doom LLM-based AI, or possibly the human race.
Consider the current state-of-the-art in "self driving cars".
They work very well... until suddenly they don't, and the meatbag behind the steering wheel gets an alert tone and has about 1 second (at most) to go from a state of distracted, book-reading, staring-vacantly-into-the-distance - to full comprehension of the situation and readiness to take corrective/evasive action. That's asking a lot of humans. I mean, I've been known to go in milliseconds from being asleep to being fully awake and vertical, but that's only when my subconscious brain hears my dog starting to retch & throw up in the corner of the room. Trust me, you don't sleep through that more than once :)
The output of LLMs runs up against the same limitation. They can be trusted X% of the time... until they can't. And the higher X goes, the more complacent we'll become, and the harder it'll bite us when they fall down and we don't spot it. "We can put AI in charge of medical diagnosis! We can put AI in charge of monitoring the skies for missile defence!" and so on.
Humans, in general, are fundamentally lazy. That's why we have spent so much of recorded history developing labour-saving devices, and it's why we're really, really bad at maintaining a state of constant alert and eagle-eyed supervision.
This post has been deleted by its author
> They're expecting people to PAY for this dross?
I pay for it... I have a rule where I will pay for one LLM at a time (so the subscriptions don't mount up).
They are great at what I'm weak at... repetitive coding tasks, coding tasks I know how to do but get bored working in my comfort zone, investment analysis (I go waaaaay deeper than LLMs, but they're a great start), and - most importantly - summarising and writing documents without sounding semi-literate.
If others don't like them, I'm fine with that. But I get fantastic benefits.
What could possibly go wrong for upstart start-ups starting in 2025 ..... other than probably everything that can go wrong because it was never ever right nor going to be right and never ever wrong?
Welcome to what is guaranteed to be an extremely exciting and uniquely challenging New Era for anything hysterical and historical and hierarchical, both conventional and traditional or novel and ground-breaking. Things aint like they used to be and there’s a great deal to learn that y’all’ll just have to accept and simply go with the flow for resistance and dissent, opposition and negative competition is futile, and very quickly realised, far too late to avoid the consequences of one’s own shared thoughts and other parties’ actions, certainly self-defeating and even deadly.
So take care out there in those Wild Wacky West and Exotic Erotic East Jungles ..... Remote Advancing Virtualised Environments. Keep IT Stupendously Simple, Do no evil if want to live and prosper in any places and spaces like there.
I hope you are using an LLM to write your posts because they are so weird, deranged and incomprehensible that I actually knew who wrote it before seeing your tag against it
Instead of The Register can I direct you towards some more suitable reading? Psychology Monthly? Therapists News? Total Nutters Digest?
I hope you are using an LLM to write your posts because they are so weird, deranged and incomprehensible that I actually knew who wrote it before seeing your tag against it itsthemonkey
Weird I might agree is acceptable, itsthemonkey, but deranged and incomprehensible I would always deny .... with future events as yet unreported supporting the rejection of what is surely only a personal opinion based upon an absence of evidence that would improve your understanding and prove your misunderunderstanding of the fact to be so.
Love your sense of humour. :-) Happy New Year to you ....... and ditto to El Reg and El Regers. GOD only knows where everything would be without yous ...... although to be perfectly honest, I think even She/He/They/IT have no clue now about what to do about yous, now that so many would be thinking themselves so AWEfully Almighty and Temporarily Invincible and Virtually Untouchable ie Dangerously Psychotic and Manifestly Delusional. ....... ie prone to constant Hallucination?
Been around a while but not really got into the comments....in fact after seeing how much this site has deteriorated over the last couple of years I will not be stopping for long,
From the fabricated "On call stories to the drivel that is "BOFH" there is very little to make up for the lack of decent technical output.
Idiots like mars man being allowed to post utter drivel without any vetting and the fact that all the columnists apparently have direct access to the true email addresses of anyone posting here are also a massive deterrent. (I know they do as I received an email from a columnist after posting a comment and was told this is their policy and they do not see anything wrong with it)
So I am only here until I find something better..or not quite as bad
Methinks thou doth protest too much, itsthemonkey. Don’t let the door hit you on the way out, but ...... if ever you think you have found something better than El Reg .... don’t be a sad ass, let all here know, so commentary can vet your decision/opinion and peer review the utility/facility for suitably appropriate improved facilities/utilities.
However, should "So I am only here until I find something better..or not quite as bad” not be true nor something readily available, are you cursed to never leave, moaning and bitching about your peers and things destined to forever remain way beyond your ken and the influence of any toxic effluent you may think to share and imagine might effect and redirect future narrative command and remote virtual control ‽
Sad is as sad does.
The quality trend of the El Reg over the last few years is a result of the general news media decline, with competition from social media reducing ad spend and forcing this site to run more intrusive ads and not be able to pay for as much content. (In my opinion, I’m not affiliated in any way).
I still find some gems here and continue to return - while I’m inclined to agree that the on call stories are a bit weak, that’s generally due to the changes in IT over the last 30 years and the aging out of the commentards, meaning the stories are now only going to work for a small niche of the readers, and are inevitably scraping the bottom of the barrel. Still, don’t bitch about the format, it’s a nice call back to reader write ins and rare to find these days. If you have any stories, why not try to formulate them into something people might be entertained by and go ahead and submit them?
As for AMFM1… congratulations on your right of passage: mistaking that account for a comprehensible debate partner. It’s not.
Happy New Year. Try not to post when you’ve been drinking, and if you weren’t, try to re-read your online posts to reduce your rage. We’ve all been there, but keep it civil, the emotion shows and it doesn’t make you any more likely to persuade.
Quite so, claimed, El Reg sure is something of a rare gem in these days where freedom of speech is maybe lauded and applauded but ruthlessly vetted, and whenever inconvenient, clumsily suppressed by shady self-serving agents and wilful third-party guardians of mainstream teleaudiovisual media operations in piratical support of a right dodgy array of historical status quo establishment arrangements/concessions/franchises, an absolutely vital practical and physical support without which they could not survive and exercise their exclusive elite executive administrative power as they whitewash unpleasant unfolding truths and paint and promote fantastical absurd pictures and future gormless notions in the hope that the ignorantly servered human will be brainwashed into believing their rantings to be the only available acceptable practical truth rather than it being recognised for what it really is .... just a desperate fiction reflecting a collection of figments of imagination catering for a few with zero regard/no due care and attention for the many.
The future however is an altogether quite different place with virtual spaces and magical environments which do not suffer the folly of fools misusing and abusing universal command and control tools and where all freely share incriminating evidence, sensitive information and secret intelligence to identify the follies of fools ignoring the aid and benefits of invaluable experience and wisdoms borne of hindsight for the extraordinary advantage rendered to the wielders absolutely fabulous fabless foresight and Almighty Insight ‽
New Year, NEUKlearer HyperRadioProACTive Environment, El Reg, with novel commentary leading with problems in need of seeds and feeds for resolving/elimination ‽ ‽ ‽
:-) Who Dares Win Wins, El Reg ...... with Special Advanced IntelAIgent Research Services for Virtual Terrain Team Territory. IT aint rocket science, is it ..... although one would have to admit it is more complicated and a great deal more dangerous than that if one doesn’t know what one is doing and what it is going to result in or where it is going to lead for the opportunities presented by such as would be a New START Treat and ACTive LOVE Threat.
Capiche?
And the result of all of that which is freely shared above is the living and excited experiencing of The Dream with AIMaster Planning with Thames Valley Crown Jewels agencies securing and protecting traders of Silicon Valley Secrets both from and for a novel assortment of noble renegade rogue hero, nutter, weirdo and misfit with odd skills aplenty to demonstrate ideally are delightful whenever not necessarily being diabolical and deadly ??
Is that where AI is currently at in the West and is it attractive and would it be helpful and rewarding to more than just all presiding and residing in the East?
And are those last four questions relating to "How to Grow Economies, at Home and Abroad, a few Qubits at a Time ... A Dummies Guide” rhetorical?
Questions, questions, questions ........ and all the time in the world to answer them as honestly as you are enabled or able.
This post has been deleted by its author
Hey ChatGPT, is it possible that AGI can not be achieved until there are further fundamental improvements in computing hardware?
Yes, it is entirely possible that Artificial General Intelligence (AGI) may not be achievable until there are further fundamental improvements in computing hardware. While progress in AI has been impressive, especially in areas like deep learning, natural language processing, and reinforcement learning, achieving AGI requires more than just better algorithms—it may also require significant advances in hardware. Here are some reasons why this could be the case: ... Energy Efficiency: Deep learning models, particularly large-scale ones, are energy-intensive. AGI may require hardware that can process massive amounts of data with orders of magnitude better energy efficiency to become practical for widespread use....
There are already profitable applications for general machine learning, including some uses of LLMs. Development while having to satisfy energy cost constraints is the secret of the evolution of life - and its no different for human progress.
Bleeding edge fundamental research is also fine - obviously - as long as honesty is maintained. To much hype is counterproductive.
FYI - According to The Information [paywalled] via Tech Crunch, Microsoft and OpenAI have a very specific, internal definition of artificial general intelligence (AGI) based on the startup’s profits ... The two companies reportedly signed an agreement last year stating OpenAI has only achieved AGI when it develops AI systems that can generate at least $100 billion in profits.This is an important detail because Microsoft loses access to OpenAI’s technology when the startup reaches AGI.
> "we thought that progress relied on key ideas produced by top researchers and that supercomputing clusters were less important" ... By 2019, it became clear that massive quantities of compute would be required...
In other words, you were not clever enough to make any new and efficient algorithms[1] that were interesting and useful to monetise, even after ransacking the literature of other people's work, never mind coming up with some ideas of your own.
But taking an old observation about an even older algorithm, a honking great pile of hardware[3] plus an enormous data shovel, a total lack of regard for publication rights and no interest in understanding how your monstrosity actually works[4], you finally managed to fill your pockets with other people's money.
So that's all right then.
[1] as the field of AI research has produced results that prove useful, even if, as soon as it is realised that we know how they work, it no longer looks like the magical thing we call our "intelligence"[2]
[2] "If it works it is not 'AI'"
[3] you know, with enough hardware, Bubble Sort can do some pretty impressive work to generate useful results for your Big Data.
[4] like, say, knowing how to direct it away from "hallucinating" or how to operate it without needing to continually add "safeguards".
Strange Game ...... only way to win is not to play. ...... Will Godfrey
...... is that a sure fire way to constantly lose against the Holy Trinity of the Strange and the Stranger and Strangest of Greater IntelAIgent Gamers, WG, and thus gravely to be regarded and summarily dismissed and publicly declared false prophetic advice and surely unhelpful information at the very least ‽
The following information from a bygone time is truly indicative of the guaranteed result of non-engagement in such as is currently at Future Building Play, and both in Work and for RESTful stateless application too ......
"Those who are not interested in politics will be forever ruled by those who are.” .......G. Edward Griffin
Damned if you do, damned if you don’t, leads and finds humanity between a rock and a hard place with nowhere else good to go, methinks.
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
