Bạn có thể thích
March 2015: Tesla CEO Elon Musk and Nvidia CEO Jen-Hsun Huang declare self-driving cars “solved” | Fortune fortune.com/2015/03/18/tes…
I just subscribed to AI Snake Oil aisnakeoil.substack.com/?utm_medium=we…
On listening, I of course also noticed which of my answers they cut (in at least one place, very obviously mid-utterance). At least one of these was a direct statement about so-called "AI" reproducing systems of oppression. Details later when I have time to relisten.
Don't take it from me (an old white man), but think for yourself: do you agree? secularhumanism.org/2022/05/identi…
#AI practitioners when faced with real-world problems other than toy ones.
Crested mynas, as many other birds, are born altricially, which means young are underdeveloped at the time of birth, therefore fed by parents. When they grow up, they have to learn that food doesn't simply jump into their beaks [📽️: Rebecca Gelernter]
I see lots and lots of that distraction. Every time one of you talks about LLMs, DALL-E etc as "a step towards AGI" or "reasoning" or "maybe slightly conscious" you are setting up a context in which people are led to believe that "AIs" are here that can "make decisions". >>
I'm not seeing any signs of that distraction "on the ground" (mostly we're making folks aware of LMs, DALL-E etc., making sure they know that the capabilities are diffusing to many actors, + discussing possible responses) but Twitter is perhaps a different story :)
Alright, fine: it's getting enough traction that I think I need to address this paper as a certified Grumpy Linguist in NLP. (I generally try to avoid peer reviewing students in public but at this point it's definitely *already* very public, so 🤷♀️.)
DALLE-2 has a secret language. "Apoploe vesrreaitais" means birds. "Contarra ccetnxniams luryca tanniounons" means bugs or pests. The prompt: "Apoploe vesrreaitais eating Contarra ccetnxniams luryca tanniounons" gives images of birds eating bugs. A thread (1/n)🧵
Scaling (i.e. "more computing") promises to become a turbid oh-how-amazing-is-AGI prompt in press releases, a bullet point more on hubris-driven tweets and conference abstracts, and the new carpet under which part of the #AI garbage will be thrown until the next buzzword is born.
ML/AI folks obsess over scale and see scale as the liberating panacea that frees ML/AI from all kinds of limitations and problems. Yet, they often fail to look into what exactly they are scaling (which is mostly toxic content as we show in this paper)
The fiddly nature of training large language models is rarely acknowledged out loud, but distilled beautifully here by @schrep, for many years CTO @ Facebook. Importantly, human children reliably learn language across wildly varying inputs.
In particular the training is fiddly and requires a lot of manual intervention. To get these models to work well for commercially relevant problems you need to fine tune them on curated datasets specific to those problems, and test them extensively against edge cases...
In Musk’s vision of Twitter and the internet, he’d be the wizard behind the curtain – projecting on the world’s screen a fake image of a brave new world empowering everyone. theguardian.com/commentisfree/…
theguardian.com
Elon Musk’s vision for the internet is dangerous nonsense | Robert Reich
Musk now faces no limit on how much of Twitter’s stock he can buy. He’ll buy as much as he needs to gain total control
this is a horrible move. AI/robots CAN'T detect lies. looks like even the EU falls for AI snake oil.
The chance that @elonmusk will deliver a robot that can come close to doing all human tasks next year is indistinguishable from zero. My comment to @CNBC: cnbc.com/2022/04/08/elo…
This week seems like win for AI, but it's actually a step back: • No info about training set [Dall-E] • Sparse disclosure of methods and errors [both] • Anecdotal data only [PaLM jokes] • Cherry-picking [both] • No access for scientific community [both]
Between OpenAI's DALL·E and Google's 540B language model, it's been a very big week for AI development news
Interested in how some of us think that the brain can help advance AI? Check out this @newscientist article where myself and fellow Neuroscientists share our views. @dendritesgr @tyrell_turing @GaryMarcus
Making a Mind: How to make machines that really think like us newscientist.com/article/mg2533…
newscientist.com
Hybrid AI: A new way to make machine minds that really think like us
In the quest to make artificial intelligence that can reason and apply knowledge flexibly, many researchers are focused on fresh insights from neuroscience. Should they be looking to psychology too?
Just 2 hours until the inaugural Irish Greenwashing Awards begin We hope you'll join us here for the live Awards thread, and we sincerely hope you'll join the conversation By working together we can address the growing greenwashing barrage in Ireland #IGA22
We are excited to announce the Irish Greenwashing Awards 2022 #IGA22 The awards are dedicated to acknowledging outstanding misinformation about the Climate Crisis across Ireland Join us this Wednesday (16th Feb) at 6pm #Greenwashing #ClimateCrisis
Some AI researchers like to read, study, think, and tend to pay attention to the meaning of the words they use. Others don't bother with any intellectual baggage and just make outrageous public statements to attract attention. Peak anti-intellectualism.
Deep learning maximalism: a dialogue A: So these guys are very confident that they will develop superhuman AI within a few years B: Cute, I guess that neuropsychology degree is really paying off for them A: No, no, they don't have any background in neuroscience or psychology
Since crypto has hit the mainstream, we've witnessed an overload of "Ponzi whataboutisms", attempts to justify crypto Ponzi schemes by comparing them to various other "things". There's an easy way to identify a Ponzi scheme...
SHOCKING: New dashcam video shows a Tesla on autopilot crash into a North Carolina state trooper's cruiser in 2020 and clip two troopers who were standing nearby after the driver was distracted while watching a movie. abc7.la/3uDN2he
United States Xu hướng
- 1. Texans 38.1K posts
- 2. World Series 113K posts
- 3. CJ Stroud 6,807 posts
- 4. Blue Jays 97.7K posts
- 5. Mariners 93.7K posts
- 6. Seahawks 37.1K posts
- 7. Springer 68.5K posts
- 8. Nick Caley 2,687 posts
- 9. White House 104K posts
- 10. StandX 4,874 posts
- 11. LA Knight 8,480 posts
- 12. Dan Wilson 4,342 posts
- 13. Sanae Takaichi 41.6K posts
- 14. Nico Collins 2,150 posts
- 15. #WWERaw 62.4K posts
- 16. Kenneth Walker 2,602 posts
- 17. Dodgers in 5 2,269 posts
- 18. Demeco 1,831 posts
- 19. Sam Darnold 4,385 posts
- 20. Baker 37.3K posts
Bạn có thể thích
Something went wrong.
Something went wrong.