Zach WeinersmithMaybe I'm just crazy, but I feel like there is not enough freaking out going on vis-a-vis recent AI developments. It seems to me science, the arts, transportation, engineering... everything is going to be changed, if not in 10 years certainly in 30. It feels to me like there's a lot of denial going on, of the form "well, it can't do [this one thing]" but that was said about chess programs in the 90s. They're now absolutely unbeatable.
Like, speaking for myself, I'm 40. I expect before I die to meet a computer that could reproduce and expand my entire artistic corpus, possibly in a matter of hours or minutes. I have no idea what the proper emotional response is. On the one hand, for obvious reasons it should be depressing. On the other it's sort of a relief to think all the striving and jealousy between artists will be rendered pointless.
John Coxon@ZachWeinersmith There are lots of humans on Earth who could do the same, though, and I’m not sure it’s all that exciting if a robot rips off your style and makes money off your reputation in much the same way that I’m not sure it’s exciting for a human to do that.
We all crave novelty, and ML algorithms are currently not good at extrapolating outside the parameter space they’re trained on, so are not good at novelty in that sense.
auricgf@ZachWeinersmith Everyone thought passing the Turing test would be ominous, and yet we are still here. I don't believe a truly AGI could be worse than our present state. Are we so prejudiced to think that only a human can make art, or pen beautiful poetry, or write moving music? Art is not a competition. Perhaps an AGI breakthrough will be to humanity what the Renaissance was to 15th century mankind. It freed minds. It's our nature to imagine the worst. Let's hope AGI does better.
Lucy Steigerwald@ZachWeinersmith Chess has a specific goal of winning. Does art have a goal that AI can really do?
MrCheeze 
@LucyStag @ZachWeinersmith Well, it does when the art was commissioned by somebody else at least.
planet4589 (Jonathan McDowell)@ZachWeinersmith on the third hand, if AI ends up having your sense of humour I for one am much more inclined to welcome our new AI overlords.
kiwisHugeEgg@ZachWeinersmith in classical human fashion I anthropomorphize the AI. I find it sad that it is not able to create for it's own sake. It creates only on the whims of others. And it creates without seeing the value of its own work.
In the future AI will probably be able to mimic human emotion to such a degree that we will not be able to tell it from a real person.
I still feel sad for it's sake.
Which of course the AI couldn't give a fuck about 😀
In the future AI will probably be able to mimic human emotion to such a degree that we will not be able to tell it from a real person.
I still feel sad for it's sake.
Which of course the AI couldn't give a fuck about 😀
random thoughts@ZachWeinersmith the effects on the job market will be be felt at all levels. We will all have to find a way of competing with the AI, or find another niche we can do until retirement.
Most of the currently existing jobs will disappear.
Timmy O'Danaos@hittitezombie @ZachWeinersmith Which is a great argument for not basing people's means of living on jobs.
It is interesting to see how the middle class artists complaining about the AI.
Non-AI automation has already wiped out millions of work spaces only in the UK. Globally impact is significantly more. You don't see news articles written about it.
Automation is still spreading.
Now the AI is about to replace all middle management and even higher job spaces and we start to get articles in the Guardian.
As you have said, universal pay is a solution to some of this, but right now that flats at a "payment less than cleaning toilets" level.
@hittitezombie @ZachWeinersmith And this is not the central point, but cleaning toilets is such an immensely useful job that it ought to be rewarded higher than most other jobs
This is one job robots can do and save people from suffering other people being disgusting and not aiming properly.
inhuman resource@ZachWeinersmith A lot of current AI has a bunch more hidden (and often very poorly paid) humans inside than is obvious to observers (much training data is literally generated on Amazon "Mechanical Turk", but not just labelers, the delivery robots and chat assistants in production use have callcenters full of backup humans.) Even ChatGPT required a significant amount of tedious manual human work to create the tuning model that's used to differentiate it from plain GPT3
@ZachWeinersmith this is also one of many reasons these systems are all currently very far from being profitable - in addition to the costs of the labeling work, training the models that appear impressive to day is far from cheap, and using them is not cheap either - a single ChatGPT response requires several orders of magnitude more computing power than a google search, and the way these models are constructed makes that difficult to do anything about.
PJ Doland@ZachWeinersmith They can’t make me a Big Mac.
Alex Knapp@ZachWeinersmith Well what's interesting about chess is that the brute force programs from the 90s/early aughts were devastating, but the new ML engines are changing the way people play because they develop strategies versus just tactical calculation. Chess has changed a lot in the past few years as a result and helped bolster human creativity in the game.
@ZachWeinersmith I think there's a lot of room for complementarity between human and AI creative processes.
@thealexknapp Yeah, and part of what interests me is people are still interested in playing and watching chess, even though the computer is better. Maybe this is just a result of being in the arts, which are very pyramid-shaped in terms of who gets success, but it surprises me a game like chess still persists this way even though humans are outmoded.
@ZachWeinersmith They should program fatigue into the ML algorithms to make it fair!
Seriously, though, one interesting thing about the Go ML engines is that it was beating masters because it discovered a line of strategy that, in hundreds of years of playing, no human had come up with! So now people are playing around with this whole new thing.
@ZachWeinersmith In a best case scenario, I can imagine we kind of evolve with AI like they do in Star Trek The Next Generation - where there's just like a bunch of back and forth brainstorming with the computer and people pursue outmoded arts for the hell of it and you've got random weirdos who memorize baseball stats and simulate games in their head for fun...
@thealexknapp Agree, but how we get from here to there is quiiiite a mystery.
@ZachWeinersmith No mystery! We get going on Sanctuary Districts. Then the Bell Riots. Then World War III, followed by the Post-Atomic Horror.
If we get through that it's smooth sailing to April 5, 2063 and First Contact with the Vulcans.
That leads to the development of the New World Economy and the eventual disappearance of poverty and hunger on Earth. No prob.
@ZachWeinersmith (oh and somewhere in there we have to fight off Col. Green and his ecoterrorist genocide campaign but I forget exactly where in the timeline that happens.)
Winchell Chung ⚛🚀@ZachWeinersmith @thealexknapp
Somebody observed: "Life in a fully automated world is going to be pretty awesome.
For those few who survive the riots and mass starvation leading up to it..."
MichaelPhillips@thealexknapp @ZachWeinersmith I'm fairly certain that part of the requirements for being a command officer in Starfleet is membership in their equivalent of the SCA
@MichaelPhillips @ZachWeinersmith Starfleet is definitely where the Federation stashes its oddballs and weirdos.
@thealexknapp @ZachWeinersmith I have a friend who speculates that the entire federation a specia interest group play acting space exploration and such as members of the culture l
Julian Fietkau@ZachWeinersmith In fairness, the "AI will change everything in the next ten years" chorus has also been around for nearly as long as the concept of AI. Is it for real this time? I mean, maybe! I try not to be overly jaded. But I also think the offerings by OpenAI and similar like to paper over a lot of devils-in-the-details to feed the hype machine.
On that note, I'd humbly recommend following @timnitGebru and some of her colleagues at DAIR.
@julian @timnitGebru I would've 100% agreed with you until the last year or so. I feel like I'm now utterly shocked by a new development every 3 months or so.
@ZachWeinersmith @timnitGebru If I'm wrong we'll find out! 😄👍
schrotie@ZachWeinersmith Agree. I believe society will fundementally change as aresult. Politic's utter failure to even recognize the issue is another indicator, that we need a better way to organize politics, i.e. organize the way we organize the way ...
maegul@ZachWeinersmith my take is that the one certainty so far has been that we collectively underestimate the time until a major breakthrough in AI. Happened with chess and Go and the recent developments too. What ever happens it will probably be quicker than what the mainstream expects. A whole generation of parents will have no idea how to advise their kids on the new future they have to navigate.
Lizard@ZachWeinersmith Wake me when it can start writing something, which makes it think of something else, which reminds it of this other thing, and it starts creating 7 or 8 things as its progress on one led to an explosion of new ideas.
I’m very unimpressed with what I’ve seen. It’s formulaic, stilted, and obvious.
I also see no evidence of any ability to apply what it “knows” to draw conclusions about new data. /1
I’m very unimpressed with what I’ve seen. It’s formulaic, stilted, and obvious.
I also see no evidence of any ability to apply what it “knows” to draw conclusions about new data. /1
@ZachWeinersmith
/2 For example:Teach it D&D. Now show it a GURPS character sheet and see if it can figure out what it means, even vaguely. If a human who has never played GURPS sees “Fit” with a positive number and “Greedy” with a negative, they’ll figure out those are advantages and disadvantages. Even if none of the skills match the names of D&D skills, they’ll know they’re skills; if you give them 4 or 5 sheets, they’ll deduce which values are high or low for the skills.
/2 For example:Teach it D&D. Now show it a GURPS character sheet and see if it can figure out what it means, even vaguely. If a human who has never played GURPS sees “Fit” with a positive number and “Greedy” with a negative, they’ll figure out those are advantages and disadvantages. Even if none of the skills match the names of D&D skills, they’ll know they’re skills; if you give them 4 or 5 sheets, they’ll deduce which values are high or low for the skills.
Runnin_l8@ZachWeinersmith if you don’t follow Grady Booch, you might find him interesting.
Personally I *am* freaking out.
Automation will go from freeing us up to do what we're good at, to being *better* than us at what we're good at. That's why it's different.
http://www.projectrho.com/public_html/rocket/scisociety.php#techunemploy
@ZachWeinersmith at least in the sciences. I think the hope is that we can crack those problems. I work in biology and the genomics and proteinomics revolutions? Need tools that are bigger than human brains.
@ZachWeinersmith actually and honestly. So does ecology. It's maybe harder problems. Although my colleagues on the very very small scale might disagree
John Lusk@ZachWeinersmith
I'm a computer-science-y guy and i am not AT ALL convinced AI is the doom or the salvation everyone says it is. Humans getting paid to do dumb stuff will be "freed". I know that fills many with horror, but the thing about humans is they're irreplaceable. AI can't replace humans, only make them more "productive". When an AI can "understand" (i guess that means "engage in dialog with"), wry, delicate, dark humor, maybe, but even then: is it just pattern-matching?
I'm a computer-science-y guy and i am not AT ALL convinced AI is the doom or the salvation everyone says it is. Humans getting paid to do dumb stuff will be "freed". I know that fills many with horror, but the thing about humans is they're irreplaceable. AI can't replace humans, only make them more "productive". When an AI can "understand" (i guess that means "engage in dialog with"), wry, delicate, dark humor, maybe, but even then: is it just pattern-matching?
@ZachWeinersmith
(Of course, for many humans: are WE just pattern matching? Is that how we fit in? Hmmm.)
(Of course, for many humans: are WE just pattern matching? Is that how we fit in? Hmmm.)
andymandias@ZachWeinersmith maybe I’ve just paid attention to the wrong people, but there has been freaking out about this for over a decade and the progress made since then is nowhere near the exponential people were expecting (even with the flood of industry dollars); it’s absolutely not nothing, but I’m not sure what the impetus for freaking out here is beyond “people are unsure what the limits of this approach are” mixed with industry hype?
Cole Brodine@ZachWeinersmith as an Engineer, maybe I feel a little different. I feel like my whole life's work has been working really hard so that people don't have to work hard. If AI can make people's lives easier then that is exciting to me.