But it doesn't seem to be any more lucid than my Markov bot, and I am still feeding it. Did someone let AMfM escape?
Most neural networks are like people with savant syndrome: they have extraordinary capabilities in a very narrow range of tasks. Now a new system built by researchers at OpenAI is more of a polymath, and has learned to perform simple language tasks without human supervision nor task-specific training. Modern neural-network …
Screams of hype, not brilliance.
Besides untrained people thought Eliza was real in 1966. The bar isn't high for a chat bot or fake news generator.
ST-TNG had Data the wrong way round. Natural English, like in certain soaps or newspapers isn't hard. Emotion simulation is easy. Playing an instrument from a score or copying art is no big deal. It's all the other autonomous everyday decisions that are impossible for AI. Is that a cat with a skin condition or a small elephant? As for actual problem solving and acceptable artistic creation? Computers can't provide decent spelling checkers or grammar correction. Yes, the word is spelled correctly but it's a completely different meaning. No, that IS correct, doubled words are not always wrong. Why do I have to look up dozens of words and add them to the "dictionary". Why doesn't the spelling checker understand context or show the definition of the suggested replacement?
Sorry, this is just hype. It's Social Media that's unsafe, not yet another text generator or chat bot.
I'm almost certain it just vaguely realised on some level that "what is" tends to be associated with looking for the "definition" of a thing, and "a dog" means the thing is "dog", so just yanked the closest thing it had to a Wikipedia article and quoted it nearly verbatim - mixing it up with some other words just enough to make sure we realize it has no idea what it is doing...
Real smart of fake smart. Either way, it sounds just like our PHB (w/o the pointy). So for now I seem to know how they did his oral output. Still wondering though how his almost human-like appearance was made. Darn, it's really not easy to distinguish from a real, human being.
Comes across like the pronouncements of certain other outfits - 'look at the wonderful thing we've done! It's better than anything ever! But no you can't have it yet, and don't look too closely at what it actually does or ask awkward questions about how'.
Maybe that stuff would fool some people but to me it still comes across as incoherent gibberish of exactly the sort you'd expect to get if you chopped up vast amounts of source material then pushed the fragments out of the grinder with some general matching for subject and some statistically driven ordering. Fucking wow!
Like a lot of modern AI it seems like a lot of resource burnt to produce something that doesn't really do much and certainly isn't intelligent. But hey, LOOK AT US!!
This thing generates random factoids that, apart from irrelevant, may be wrong or very wrong? And it "tends to wander off incoherently after a few paragraphs"?
Man, this is going to save me so much time! I'll be able to delegate all my internet posting to this machine. Where do I get a licence?
A piece of lame software spewing idiotic nonsense and they claim "software is fascinating and impressive from a technological and academic point of view"! Gosh, those academics are so easily impressed that I'm truly scared by the future they envision for human kind.
Article: OpenAI says it is withholding portions of the software to prevent it being abused to automatically churn out the equivalent of deep fake videos for the written word
Judging by the samples, a very, very premature concern. A human would do a much better job of faking someone's writing. In fact, this "AI" is very far from fooling anyone who reads more than a few paragraphs of the output.
Flak for Slack chaps in yak app hack flap: User database whacked
El Reg Headlines are often poetry and require an understanding of humour and much deeper language skills.
I'm not surprised that the poor AI engine needed a Solpadeine and a lie-down. You were asking it to get meta.
Maybe a Gumby brothers icon ("My brain hurts! my brain hurts!").
YouTube clip for those who have temporarily forgotten it: https://www.youtube.com/watch?v=c7AihnLbw1E
I agree that this bot is dangerous, but not for the reasons stated. The content it produces is garbage, even when it's written in understandable English. Could it be confused by something written by a human? Sure, just read how all those NPC are regurgitate their party-line. It's hard on the same level.
With this bot spamming forums and discussions until any semblance of content is gone will be easy. For the moderators, it'll become a nightmare to sort out your average angry internet-citizen with too many pet-peeves from a bot. Keeping up a no-censorship policy will become even harder if you still want to keep useful content.
The icon because it'll make flamewars even easier.
Perhaps before we start drawing up Asimovian rules for AI, such as "a robot must never lie to a human", we should teach them the difference between fact and fiction! Incidentally, how did El Reg play with this thing? I smell an opportunity... Train the wordy bot on my past catalog of emails and ticket notes, let it loose on my PC, and sneak out early for a ---------->
GPT2 will not be able to hold a candle to anything or anyone if serially led by heads of policy without a clue about what to do about what can be so easily done nowadays.
While GPT2 is potentially groundbreaking, OpenAI is not ready to share it with the world just yet. Its head of policy told the Guardian that GPT2 needs to be tinkered with for a while to find out whether it could be used for mischief. .... AI firm is keeping a text generating tool under wraps amid fears
Whether it could be used for mischief? Oh, I was thinking much more about however do you limit and/or mitigate mischievous use, ignorant misuse and arrogant abuse. And how to efficiently terminally punish and illiquidate the crazily dangerous miscreant in cases of constant abuse and perfectly understood misuse to extremis and beyond.
Sound safe advice from all of that? ...... Don't be a crazily dangerous miscreant, for enemies are myriad and all-consuming/stealthy and designedly deadly when no prisoners are required.
I think you shoud take a stress pill amfM ... .....Cliff Thorburn
What on Earth for, Cliff T, whenever everything is going so perfectly well. :-) ..... Deeply Embedded into SCADA Systems and MetaDataBase Mining and Prospecting with COSMIC Partners.
And be assured, Heavenly is not Stressful :-) Extremely Exciting Definitely yes and Never Overwhelming to That and Those of the Ken is the Right Devil of a Temptation to Satisfy with Constant Success for Entry into that Supreme Ancient Sublime Program.
As you will have to imagine, that is one helluva journey to travel always by all ways made available.
Glad to hear :-)
After another depressing Saturday, what I like to refer to as ‘Psy Ops Live Operational Virtual Environments’ curtailing Valentines Day, but not showing much love mostly hate, I often wonder if the sheeple actually are concious of their concerted activities, or the true hands who rock the cradle.
Think I may pop a stress pill myself, and contemplate whether humanity will ever return to more aspirational centrist morality instead of dancing round the maypole of tortious criminality.
I often wonder if the sheeple actually are concious of their concerted activities, or the true hands who rock the cradle. .... CT
Methinks the default virtual state for practically everything but a blessed and/or cursed few is Blissful Ignorance, CT .... which is probably just as well for whenever the masses be told, and it cannot be disproved, that they have been used and abused as the expendable playthings of an Exclusive Executive Administrative Operating System, with levers pushed and pulled by a Corrupted and Perverse Elite, will there be No Secure Hiding Place and No Safe Haven for the Worthy of Retribution and Just Merciless Revenge/Merciful End.
Capiche, El Reg? Are you working out what has already been done and where everything is going?
Words Create, Command and Control and Destroy Cruel Worlds in a Flash with a Crash of Intelligence Systems Revealing 0Day Vulnerabilities to Exploit and Explore/Wield and Monetize especially if of particular and peculiar interest to you and any following supportive partners/Angel Investors.
:-) For AI, but not as you were expecting it to be in IT.
“Capiche, El Reg? Are you working out what has already been done and where everything is going?”
I’m pretty sure El Reg regulars, and even the not so regular are intelligent enough to understand that many, much and most of the worlds, and Brexit woes are much more than the SIMple decision of what mobile network to select, and more akin to catastrophic failures and corroborated corruption elsewhere down the pipeline, and failure to accept responsibility for such.
Ministers of War, manic panic, and failure to uphold the very virtues proclaimed at the very heart of the West and Midwest’s best.
I’m pretty sure El Reg regulars, and even the not so regular are intelligent enough to understand that many, much and most of the worlds, .... Cliff Thorburn
Oh? As much as that would be nice and exciting, I'm pretty sure they are not, given the fact that they are so subservient and apathetic to serial abuses by ....... well, they would be that and those whose identifier be membership and/or unquestioning support of the powers that be [TPTB] both politically adept and inept, corrupt and subversive.
Things however are rapidly changing, and evolving and mutating in the rancid theatres of FUD and SNAFU ........
“The whole liberal world order appears to be falling apart – nothing is as it once was." Wolfgang Ischinger, chairman of the Munich Security Conference, wrote in an essay introducing the conference. “We are experiencing an epochal shift; an era is ending, and the rough outlines of a new political age are only beginning to emerge. No matter where you look, there are countless conflicts and crises"......... and such progress in unstoppable.
Would there be any/many who would disagree and posit such as nonsense and fake news reported and in so doing, prove the initial uncomfortable assertion above true and a fact afloat in seas and sees of fiction?
This post has been deleted by its author
Biting the hand that feeds IT © 1998–2021