El fin del trabajo

[Saturio]

https://techcrunch.com/2025/01/08/elon-musk-agrees-that-weve-exhausted-ai-training-data/

Elon Musk concurs with other AI experts that there’s little real-world data left to train AI models on.

“We’ve now exhausted basically the cumulative sum of human knowledge … in AI training,” Musk said during a livestreamed conversation with Stagwell chairman Mark Penn on X late Wednesday. “That happened basically last year.”

Musk, who owns AI company xAI, echoed themes former OpenAI chief scientist Ilya Sutskever touched on at NeurIPS, the machine learning conference, during an address in December. Sutskever, who said the AI industry had reached what he called “peak data,” predicted a lack of training data will force a shift away from the way models are developed today.

Indeed, Musk suggested that synthetic data — data generated by AI models themselves — is the path forward. “The only way to supplement [real-world data] is with synthetic data, where the AI creates [training data],” he said. “With synthetic data … [AI] will sort of grade itself and go through this process of self-learning.”

Other companies, including tech giants like Microsoft, Meta, OpenAI, and Anthropic, are already using synthetic data to train flagship AI models. Gartner estimates 60% of the data used for AI and an­a­lyt­ics projects in 2024 were syn­thet­i­cally gen­er­ated.

Microsoft’s Phi-4, which was open sourced early Wednesday, was trained on synthetic data alongside real-world data. So were Google’s Gemma models. Anthropic used some synthetic data to develop one of its most performant systems, Claude 3.5 Sonnet. And Meta fine-tuned its most recent Llama series of models using AI-generated data.

Training on synthetic data has other advantages, like cost savings. AI startup Writer claims its Palmyra X 004 model, which was developed using almost entirely synthetic sources, cost just $700,000 to develop — compared to estimates of $4.6 million for a comparably sized OpenAI model.

But there as disadvantages as well. Some research suggests that synthetic data can lead to model collapse, where a model becomes less “creative” — and more biased — in its outputs, eventually seriously compromising its functionality. Because models create synthetic data, if the data used to train these models has biases and limitations, their outputs will be similarly tainted.

Pues, vaya, a pesar de eso, los generadores de imágenes siguen sin saber cómo funciona un reloj (o lo que es un reloj o lo que es el tiempo).
Prueben a pedirles una imagen de un reloj (analógico) señalando las cuatro y media.

 

[Cadavre Exquis]

https://x.com/ggerganov/status/1894057587441566081

Saludos.

[muyuu]

https://x.com/MarioNawfal/status/1994488967379529914

🇪🇸🇺🇸 SPAIN HANDED TESLA THE KEYS TO EUROPE: UNLIMITED NATIONWIDE FSD TESTING, NO DRIVER REQUIRED

Yes, this is massive news for Tesla.

Spain quietly dropped the mother of all regulatory gifts in July 2025: the ES-AV framework that puts the country straight into Phase 3.

Remote monitoring allowed, no mandatory safety driver, full public road access.

Tesla immediately got approval for 19 vehicles with unlimited testing across the entire country.

This isn't some small pilot program in a parking lot.

This is real-world, all-roads, no-human-behind-the-wheel testing at scale, exactly what Tesla needs to train FSD Supervised and Robotaxi to European driving chaos.

Spain just became Tesla's European data goldmine overnight.

While California is still choking on red tape and Germany moves at a bureaucratic snail's pace, Spain said, "Come get your miles."

2026 Robotaxi fleets in Madrid and Barcelona suddenly look a lot more real.

Is Europe waking up? If so, Tesla is the alarm clock.

Source: ES-AV Framework Programme (July 2025),
@KRoelandschap
,
@Tesla
,
@teslaeurope

[muyuu]

https://www.youtube.com/watch?v=-m9kY1cpmgw

Kissinger and the Future of AI ft. Eric Schmidt


39:16 (parte)

00:39:20.461 --> 00:39:24.138
THE MOST, YOUR LINE OF SIGHT.
≫ WHY IS THIS MADNESS

00:39:24.138 --> 00:39:26.138
OCCURRING?

00:39:26.944 --> 00:39:29.759
IT MUST BE A BUBBLE AND IT'S
GOING TO CRASH. NO. IT'S NOT A

00:39:29.759 --> 00:39:31.759
BUBBLE.

00:39:33.066 --> 00:39:35.066
IF ANYTHING IT'S UNDER HYPED
BECAUSE YOU ARE

00:39:36.274 --> 00:39:38.274
FUNDAMENTALLY AUTOMATING
BUSINESSES AND THE REASON

00:39:38.274 --> 00:39:40.274
PEOPLE ARE SPENDING THIS AMOUNT
OF MONEY IS TO AUTOMATE THE

00:39:40.275 --> 00:39:43.391
BORING PARTS OF  THEIR
BUSINESS. WHETHER IT'S BILLING

00:39:43.391 --> 00:39:45.391
OR ACCOUNTING

00:39:45.994 --> 00:39:48.704
OR PRODUCT DESIGN OR DELIVERY
OR INVENTORY OR WHATEVER.

00:39:48.704 --> 00:39:50.704
PEOPLE ARE AUTOMATING THOSE.
AND THERE'S AN AWFUL LOT THERE.

00:39:51.513 --> 00:39:54.622
THINK ABOUT MEDICINE. CLIMATE
CHANGE AND ENGINEERING NEW

00:39:54.622 --> 00:39:56.622
SCIENCE.

00:39:59.538 --> 00:40:01.538
IT'S EXTRAORDINARY.

00:40:06.945 --> 00:40:10.553
≫ WHAT EXCITES YOU THE MOST OF
THE ONES  THAT YOU'VE SEEN THAT

00:40:10.553 --> 00:40:14.861
YOU HAVE A LINE OF SIGHT TO
THAT  THE REST OF US ARE  NOT

00:40:14.861 --> 00:40:16.861
SEEING.
≫ SORRY.

00:40:16.861 --> 00:40:18.861
I HAVE A COUGH.

00:40:22.103 --> 00:40:24.103
I APOLOGIZE.

00:40:26.325 --> 00:40:29.337
≫ WE CAN ALL SEE IN OUR OWN
IMAGINATION WHAT WE ARE

00:40:29.337 --> 00:40:31.337
THINKING OF BUT WE WILL SEE
WHAT ERIC SAYS.

00:40:32.646 --> 00:40:34.646
≫ WHEN I STARTED IN HIGH SCHOOL
I WAS AN

00:40:36.254 --> 00:40:40.461
EARLY PROGRAMMER AND I
DELIGHTED IN  WRITING CODE.

00:40:40.461 --> 00:40:42.461
WHEN I WENT TO COLLEGE AND
GRADUATE SCHOOL THAT'S ALL I

00:40:42.461 --> 00:40:45.570
WANTED  TO DO. I WAS THE
DEFINITION OF A NERD AT  THE

00:40:45.570 --> 00:40:49.071
TIME. AND EVERYTHING THAT I DID
IN MY 20s WHICH GOT ME TO WHERE

00:40:49.071 --> 00:40:51.071
I AM HAS NOW BEEN  COMPLETELY
AUTOMATED.

00:40:54.386 --> 00:40:56.386
EVERY ASPECT OF THE PROGRAMMING
I DID, EVERY ASPECT OF DESIGN

00:40:56.386 --> 00:40:58.497
IS NOW DONE  BY COMPUTERS.

00:41:01.702 --> 00:41:04.624
I RECENTLY HAD IT RIGHT A WHOLE
PROGRAM  FOR ME. WATCHING IT

00:41:04.624 --> 00:41:06.624
DEFINE THE CLASSES AND DETAIL
OF THE INDIRECTION AND SO

00:41:06.624 --> 00:41:10.680
FORTH. HOLY CRAP. THE END OF
ME.

00:41:13.894 --> 00:41:15.894
AND I'VE BEEN DOING PROGRAMMING
FOR

00:41:17.154 --> 00:41:20.971
55 YEARS. SO TO SEE SOMETHING
START AND END IN FRONT OF

00:41:23.980 --> 00:41:25.980
YOUR OWN LIFE IT'S REALLY
PROFOUND.

00:41:26.807 --> 00:41:30.122
COMPUTER SCIENCE IS NOT GOING
AWAY. AT LEAST UNTIL COMPUTER

00:41:30.122 --> 00:41:35.845
SCIENTIST GETS REPLACED WILL BE
SUPERVISING THIS. BUT THE

00:41:35.845 --> 00:41:38.366
ABILITY TO GENERATE CODE AT THE
POWER  THAT THESE SYSTEMS CAN

00:41:38.366 --> 00:41:40.366
DO IS REVOLUTIONARY.

00:41:42.471 --> 00:41:44.471
EACH AND EVERY ONE OF YOU HAS A
SUPERCOMPUTER AND SUPER

00:41:44.471 --> 00:41:46.471
PROGRAMMER IN  YOUR POCKET.

00:41:47.263 --> 00:41:50.166
NOBODY HERE IS  A TERRORIST.
IT'S ALWAYS EASIER TO USE

00:41:50.166 --> 00:41:53.476
NEGATIVE EXAMPLES. THERE'S
PLENTY OF I WILL USE A

00:41:53.476 --> 00:41:55.476
STEREOTYPE, YOUNG MEN LIVING IN
THE

00:41:56.681 --> 00:41:58.681
BASEMENT, THEIR MOTHERS GIVE
THEM FOOD AND

00:42:03.488 --> 00:42:05.488
BASIC THEIR -- THEY SIT THERE.

00:42:08.416 --> 00:42:10.416
THEY ALL HAVE THE ABILITY TO
USE THESE TOOLS TO BUILD

00:42:10.416 --> 00:42:13.229
INCREDIBLY POWERFUL SYSTEMS.
CYBER ATTACKS,  OTHER THINGS.

00:42:16.447 --> 00:42:18.447
THERE'S SOME EVIDENCE

00:42:21.261 --> 00:42:23.261
THAT THE FELLOW WHO KILLED THE
INSURANCE EXECUTIVE WAS INTO

00:42:23.261 --> 00:42:25.273
SOME  OF THIS, PEOPLE WERE
LOOKING AT SOME OF HIS

00:42:25.273 --> 00:42:29.370
WRITINGS. OF COURSE HE IS IN
JAIL NOW BUT THAT HE WAS

00:42:31.073 --> 00:42:33.385
SOMEHOW INFLUENCED. IT'S AN
EXAMPLE OF SOME OF THE DARKEST

00:42:33.385 --> 00:42:35.385
RECESSES OF HUMANITY.

00:42:37.099 --> 00:42:39.099
YOU GIVE THOSE PEOPLE THESE
KINDS OF TOOLS, WE HAVE TO BE

00:42:39.099 --> 00:42:41.099
READY. THE INDUSTRY IS WELL
AWARE OF THIS AND WORKING ON

00:42:41.099 --> 00:42:43.497
IT. IT'S VERY IMPORTANT THAT
DEFENSIVE SYSTEMS BE CAPABLE.

00:42:46.716 --> 00:42:51.046
THE EVENTUAL SOLUTION  TO A.I. ,
GOOD A.I. FIGHTING BAD A.I..

00:42:53.266 --> 00:42:55.266
THAT'S HOW IT WILL RESOLVE
ITSELF.

00:42:56.682 --> 00:42:59.302
≫ I WANT TO ASK YOU ABOUT THE
U.S. CHINA RIVALRY IN A.I.

00:43:02.508 --> 00:43:04.927
AS YOU  SEE IT. APOLOGIES IF
IT'S NOT CLEAR ENOUGH.

00:43:08.130 --> 00:43:10.130
BUT IT SUGGESTS IF WE TAKE A
SERIES

00:43:13.135 --> 00:43:15.135
OF INDICES, IF YOU LOOK AT THE
PERFORMANCE GAP IN

00:43:16.247 --> 00:43:19.459
JANUARY OF 24 IT WAS
SIGNIFICANTLY LARGER THAN IT

00:43:21.970 --> 00:43:23.970
IS TODAY.

00:43:26.687 --> 00:43:28.893
AND WHAT WE MAKE OF THIS AND
WHAT WE MAKE OF THE LIKELY

00:43:28.893 --> 00:43:30.893
FUTURE?

00:43:31.917 --> 00:43:33.917
≫ THE CHART IS CORRECT.

00:43:36.785 --> 00:43:38.785
BUT THE PEOPLE WHO ARE
INFLUENCED BY THIS CLAIM THAT

00:43:38.785 --> 00:43:40.785
IT'S NOT GOING TO BE  TRUE FOR
LONG.

00:43:42.695 --> 00:43:44.695
BECAUSE THE REASONING
REVOLUTION REQUIRES SO MANY

00:43:44.695 --> 00:43:46.695
CHIPS AND SO MUCH OF THE  MAGIC
THAT THE SAN FRANCISCO PEOPLE

00:43:46.695 --> 00:43:50.830
HAVE INVENTED, USING THAT  AS A
SORT OF MONIKER. THAT THE GAP

00:43:50.830 --> 00:43:54.144
WILL WIDEN. MY OWN VIEW IS THAT
THE GAP WILL WIDEN BUT FOR

00:43:54.144 --> 00:43:56.144
DIFFERENT REASONS.

00:43:57.453 --> 00:43:59.453
I THINK THAT THE CHINESE FOCUS
IS LARGELY AS I MENTIONED  ON

00:43:59.453 --> 00:44:01.453
EMBEDDING A.I.

00:44:02.570 --> 00:44:05.975
IN EVERYTHING. TOASTERS, CARS,
THEY ARE MOVING MUCH

00:44:06.882 --> 00:44:10.284
MORE QUICK. THE VAST MAJORITY
OF HUMAN ROBOTS WILL BE

00:44:13.296 --> 00:44:15.296
CHINESE AND HIGH-POWERED AND

00:44:16.396 --> 00:44:18.396
MANUFACTURED BECAUSE THEY KNOW
HOW TO DRIVE THE COST OF THINGS

00:44:18.396 --> 00:44:20.396
DOWN. THE SUPPLY CHAINS ARE
INCREDIBLE, THE COST

00:44:20.396 --> 00:44:24.020
MANAGEMENT, ALL OF THE
DIFFERENT STUFF. SO MY GUESS IS

00:44:24.020 --> 00:44:26.020
THAT IT'S

00:44:27.226 --> 00:44:29.226
TRUE THAT THE GAP WILL PROBABLY
GET

00:44:31.831 --> 00:44:33.831
LARGER BUT THE REAL QUESTION IS
WILL YOU AS A CONSUMER

00:44:33.831 --> 00:44:35.831
ULTIMATELY HAVE A BETTER
EXPERIENCE WITH A CHINESE

00:44:35.831 --> 00:44:39.651
PRODUCT THAN A U.S. PRODUCT.
THE ANSWER IS PROBABLY THE

00:44:39.651 --> 00:44:41.651
CHINESE PRODUCT AND THAT'S OF

00:44:41.860 --> 00:44:43.860
CONCERN.

00:44:48.698 --> 00:44:50.698
≫ THERE IS A HALF-DOZEN
QUESTIONS ABOUT WHICH PEOPLE

00:44:50.698 --> 00:44:53.513
ARE MAKING BETS AND YOU'VE
THOUGHT ABOUT IT DEEPLY. ARE WE

00:44:53.513 --> 00:44:55.513
GOING TO BET ON

00:44:57.923 --> 00:44:59.923
COMPUTER CHIPS AND SNACKS  OR
BRAINS?

00:45:01.632 --> 00:45:04.934
CLOSED OR OPEN? ANOTHER ONE IS
ARE WE GOING TO BET WORKING

00:45:04.934 --> 00:45:06.934
HARD

00:45:08.135 --> 00:45:10.244
ON AGI  OR APPLICATIONS?

00:45:13.446 --> 00:45:15.446
IF YOU GO ACROSS THE

00:45:17.251 --> 00:45:19.251
SPECTRUM HERE, IF I LOOK AT THE
CHINESE

00:45:21.256 --> 00:45:25.767
PIECE CERTAINLY THE GUYS  AT
DEEP-SEA HAVE 200 PEOPLE WHO

00:45:29.075 --> 00:45:31.075
HAVE BRAINS THAT COST

00:45:32.082 --> 00:45:35.492
ONE 1000 OF THE COST OF A.I.
AND NOW THERE ARE

00:45:37.806 --> 00:45:40.316
SIX OTHER ALONG FROM THE SAME
SPACE. SO THAT ONE MAKES ME

00:45:40.316 --> 00:45:43.726
WARY. ON THE CLOSED VERSUS
OPEN, OUR

00:45:47.434 --> 00:45:49.434
LAST CONVERSATION YOU WERE
PRETTY MUCH

00:45:52.243 --> 00:45:54.243
CONCLUDED THAT OPEN IS GOING TO
BE CLOSED BUT ALL OF OUR

00:45:54.243 --> 00:45:56.243
COMPANIES ARE MOSTLY CLOSED. SO
WHAT ABOUT THAT?

[Cadavre Exquis]

Sal Khan: Companies Should Give 1% of Profits To Retrain Workers Displaced By AI
Posted by EditorDavid on Sunday December 28, 2025 @03:37AM from the stopping-a-job dept.

"I believe artificial intelligence will displace workers at a scale many people don't yet realize," says Sal Kahn (founder/CEO of the nonprofit Khan Academy). But in an op-ed in the New York Times he also proposes a solution that "could change the trajectory of the lives of millions who will be displaced..."

"I believe that every company benefiting from automation — which is most American companies — should... dedicate 1 percent of its profits to help retrain the people who are being displaced."

This isn't charity. It is in the best interest of these companies. If the public sees corporate profits skyrocketing while livelihoods evaporate, backlash will follow — through regulation, taxes or outright bans on automation. Helping retrain workers is common sense, and such a small ask that these companies would barely feel it, while the public benefits could be enormous...

Roughly a dozen of the world's largest corporations now have a combined profit of over a trillion dollars each year. One percent of that would create a $10 billion annual fund that, in part, could create a centralized skill training platform on steroids: online learning, ways to verify skills gained and apprenticeships, coaching and mentorship for tens of millions of people. The fund could be run by an independent nonprofit that would coordinate with corporations to ensure that the skills being developed are exactly what are needed. This is a big task, but it is doable; over the past 15 years, online learning platforms have shown that it can be done for academic learning, and many of the same principles apply for skill training.

"The problem isn't that people can't work," Khan writes in the essay. "It's that we haven't built systems to help them continue learning and connect them to new opportunities as the world changes rapidly."

To meet the challenges, we don't need to send millions back to college. We need to create flexible, free paths to hiring, many of which would start in high school and extend through life. Our economy needs low-cost online mechanisms for letting people demonstrate what they know. Imagine a model where capability, not how many hours students sit in class, is what matters; where demonstrated skills earn them credit and where employers recognize those credits as evidence of readiness to enter an apprenticeship program in the trades, health care, hospitality or new categories of white-collar jobs that might emerge...

There is no shortage of meaningful work — only a shortage of pathways into it.

Thanks to long-time Slashdot reader destinyland for sharing the article.

Saludos.