SoftBank CEO says AI that is 10,000 times smarter than humans will come out in 10 years

erek

erek

[H]F Junkie
Joined
Dec 19, 2005
Messages
11,306
Re: SoftBank shares swept up in AI chip frenzy
If you’re liking the Arm IPO you'll love the RISC-V IPO “SoftBank CEO Masayoshi Son, who has...

"Son laid out his vision for a world featuring artificial super intelligence, or ASI, as he dubbed it."

1719192617568.png


"Son discussed how the future will hold various ASI models that interact with each other, like neurons in a human brain. This will lead to AI that is 10,000 times smarter than any human genius, according to Son.

SoftBank shares closed down more than 3% in Japan, following the meeting.




Son is SoftBank’s founder, who rose to prominence after an early and profitable investment in Chinese e-commerce giant Alibaba. He positioned SoftBank as a tech visionary with the 2017 launch of the Vision Fund, a massive investment fund focused on backing tech firms. While some of the bets were successful, there were also many high-profile failures, such as office sharing company WeWork.

After posting then-record financial losses at Vision Fund in 2022, Son said that SoftBank would go into “defense” mode and be more conservative with its investments. In 2023, the Vision Fund posted a new record loss, with Son shortly after saying that SoftBank would now shift into “offense,” because he was excited about the investment opportunities in AI.

Son has been broadly out of the public eye since then.

He returned to the spotlight on Friday to deliver a speech that was full of existential questions.

“Two years ago, I am getting old, rest of my life is limited, but I haven’t done anything yet and I cried so hard,” Son said, suggesting he feels he hasn’t achieved anything of consequence to date.

He added that he had now found SoftBank’s mission, which is the “evolution of humanity.” He also said he has discovered his own purpose in life.

“SoftBank was founded for what purpose? For what purpose was Masa Son born? It may sound strange, but I think I was born to realize ASI. I am super serious about it,” Son said."

Source: https://www.cnbc.com/2024/06/21/sof...that-is-10000-times-smarter-than-humans-.html
 
Anyone who believes that AI will be as smart as a human, let alone 10,000 times smarter is the biggest idiot. Then again this statement was meant to appeal to investors so job well done. The is the problem with calling Machine Learning AI, because people believe these algorithms can think. Especially with a prediction that's 10 years off, which you can stick this with goals like 100% renewables and fusion energy. Softbank must be ready to file for bankruptcy to say something so bold and unrealistic.
 
Sure, if people are so stupid that they can't figure out how to use it in 10 years. Seems plausible.
 
So to recap:

2019: Sold all Nvidia shares for $3.9B, that would've been worth 50x today if he'd held
2022: Record financial losses
2023: Even bigger financial losses
2024: "Predicts" 10000x smarter AI in 10 years, because everyone knows things happen in big round numbers and multiples of ten

And he talks about himself in the third person. CEO of the decade candidate.
 
Last edited:
DPI said:
So to recap:

2019: Sold all Nvidia shares for $3.9B, that would've been worth 50x today if he'd held
2022: Record financial losses
2023: Even bigger financial losses
2024: "Predicts" 10000x smarter AI in 10 years, because everyone knows things happen in big round numbers and multiples of ten

CEO of the decade, ladies and gentlemen.
Click to expand...
Also the massive wework investment losses

1719287807949.png
 
Shoganai said:
AI isn't even "smart" to begin with. It's not sentient. It's not even AI.
Click to expand...
Would it be actually smart instead mimicking it, wouldn't it stop to be artificial ?

DukenukemX said:
Anyone who believes that AI will be as smart as a human, let alone 10,000 times smarter is the biggest idiot.
Click to expand...
Anyone that believe anything about this sound like someone that completely overrated their knowledge and ability to read the futur. You think it is impossible that in the year 3,000 the best computer system in the world will not be with a reasonable definition of smart and test for it, be as smart as an average human of today ?

Yesterday people were convinced machine would never be able to play chess, we obviously have not the smallest idea how smart a synthetic system can become (and would have a really hard time imagining what it would be like for something to be smarter than us, almost by definition), it is almost impossible you believe what you just wrote.

DukenukemX said:
The is the problem with calling Machine Learning AI, because people believe these algorithms can think.
Click to expand...
Now we are back from AI will destroy profession to algorithms....
 
They call it AI, but it's really Machine Learning, which is NIFTY. but we already have algorithms that can exceed 10,000x the capability of a human. That's what computers do.

Try finding the square-root of a triple-digit number to the nearest hundredths by hand. I would say a computer is going to be WAY more than 10,000x faster than any human.
 
I wonder how one define smart (could be a translation issue, was it in english ?) to even make a statement about something being 10,000 smarter than something else, achieving the same solution 10,000 faster, would it make it 10,000 smarter, probably not. In 10 years, bet are good it will be inhumain smarter and knowledge filled than humans in many ways and still inhumanly dummer in many ways (like it is already)

Humans tend to call human smart for stuff computer can do, it becomes more being a parlor trick than smart if your brain is good at something a computer also good at, in the past being good at playing chess some would have called smart, being good at making complicated mental calculus some would have called it smart, we will probably continue to do so
 
LukeTbk said:
Anyone that believe anything about this sound like someone that completely overrated their knowledge and ability to read the futur.
Click to expand...
Careful, it's learning. You seem to agree with Softbank, so explain how they'll make something 10,000 times smarter than a human?
You think it is impossible that in the year 3,000 the best computer system in the world will not be with a reasonable definition of smart and test for it, be as smart as an average human of today ?
Click to expand...
Last I checked the year 3,000 is not 10 years from now. Unless I'm developing Alzheimer's in which case what year is this? Give enough time and we'll cure aging but Softbank said 10 years not 1,000 years.
what-year-is-it-robin-williams.jpg

Yesterday people were convinced machine would never be able to play chess, we obviously have not the smallest idea how smart a synthetic system can become (and would have a really hard time imagining what it would be like for something to be smarter than us, almost by definition), it is almost impossible you believe what you just wrote.
Click to expand...
Meanwhile we had chess games on the NES called Battle Chess that had an AI you could play against. Yes IBM's chess AI is leagues better but nobody cares. Impressing idiots is not an accomplishment, which is exactly what Softbank is trying to do with that statement. Softbank wouldn't be high on my list for companies that could pull it off. All we'll be able to do in 10 years is create a very convincing BonziBuddy. Maybe this time it won't recommend glue on Pizza.
330px-Bonzi_Buddy.png

Now we are back from AI will destroy profession to algorithms....
Click to expand...
I already explained this to you with the dot com boom as an example. Something will come out of all this AI work, but not in a way most people would expect. Mostly because nobody has made any cool ideas with AI and what ideas they have created are limited by the lack of accuracy from the AI. I'm hoping we can make good use out of these TPU's instead of the Copilot and Recall crap Microsoft is pushing down peoples throats.
KazeoHin said:
They call it AI, but it's really Machine Learning, which is NIFTY. but we already have algorithms that can exceed 10,000x the capability of a human. That's what computers do.
Click to expand...
We have machines that can exceed humans in a very specific simple task. Machine Learning is trying to push this towards more complicated tasks that humans do with ease, like talking and driving. Even then it's just mimicking it, and not actually learning. Like if I'm talking to AI and use a new slang word like rizz, it's not something it could understand with context. Most people would understand what it means even if they've never it before. AI is just gonna assume it's a mispronunciation and find a word that sounds similar.
Try finding the square-root of a triple-digit number to the nearest hundredths by hand. I would say a computer is going to be WAY more than 10,000x faster than any human.
Click to expand...
Now tell it to drive a car down a road that's nothing but rocks and weeds. Intelligence isn't something that can be measured with a single act.
 
Ok so, assuming his prediction here is as good as his skills at predicting what is good in the market in 10 years we will have AI that is worse than a 1990s chatbot and can only communicate in Klingon.
 
Sycraft said:
Ok so, assuming his prediction here is as good as his skills at predicting what is good in the market in 10 years we will have AI that is worse than a 1990s chatbot and can only communicate in Klingon.
Click to expand...
I mean, ML LLMs are primarily being used for Social Media posting provocative or targeted AI generated images, Bots are taking over most of the online world to the point that I wouldn't be surprised if in 10 years time people with any amount of agency in their life mostly move away from social or internet media entirely, so yeah, AI will be so convincing at targeting ragebait or walmart/shein ads to your mom or dad it'll have no real-world use as everyone just avoids avenues where it exists, becoming functionally useless.
 
DukenukemX said:
Careful, it's learning. You seem to agree with Softbank, so explain how they'll make something 10,000 times smarter than a human?
Click to expand...
Saying no one knows about the future is agreeing about someone telling us the future ? All the big name in interviews (the DeepMind founder, the Sam Altman), all seem to say how much we do not know what it will look like in the near future let alone what the physical law of nature limits would be.

DukenukemX said:
Last I checked the year 3,000 is not 10 years from now. Unless I'm developing Alzheimer's in which case what year is this? Give enough time and we'll cure aging but Softbank said 10 years not 1,000 years.
Click to expand...
What does 10 years or 1000 years has to do with the statement: Anyone who believes that AI will be as smart as a human, it is not like there any time limit to it and with the speed of change in the frontier tech 10 years is already a wild future (we can predict giant inertia, laws, fears, could make simply fully integrating a GPT-5 capability that run fast and super cheap, just that could take 30 years, but in terms of what run inside google Lab, we are maybe 2 attention is all you need transformer jump to what exactly, who knows).

Robots trained in a physical world that interact with, would have way more sense than human (we are limited to around 11-13 major one, well up to 21 depending on the definition) and their simulated version in simulation of the world, will be able to be exposed to inhuman amount of experience to "learn" from.

DukenukemX said:
Meanwhile we had chess games on the NES called Battle Chess that had an AI you could play against.
Click to expand...
Yes Blaise Pascal and other prediction that never a machine could play chess when he was starting to make calculating machine were proven wrong a while ago, the exact point of people trying to predict the future of synthetic intelligence being extremely hard, about every genius in history was wrong about it, what do you think make you specially good at knowing ? Do you believe there anything magic about an human brain that would make it physically impossible for a synthetic system to do everything it does and much more ?

There is yet to have any reason to believe that anything in the human mind cannot be done in a synthetic one.

DukenukemX said:
I already explained this to you with the dot com boom as an example. Something will come out of all this AI work, but not in a way most people would expect. Mostly because nobody has made any cool ideas with AI and what ideas they have created are limited by the lack of accuracy from the AI. I'm hoping we can make good use out of these TPU's instead of the Copilot and Recall crap Microsoft is pushing down peoples throats.
Click to expand...
If really mRNA vaccine development, Alpha-fold 3, a new open source CRISPR gene editing protein, predicting the meteo 10 days in any gps location on earth better than any super computer are not cool things, what would it be, self-driving car and above ? I use Github copilot pretty much everyday and you need to pay for it, no one had it push down their throats.

If AI is the internet of 1994 and anything similar to the dot com boom it will be the biggest thing ever, but what the link to say that algorithm (a process or set of rules to be followed in calculations or other problem-solving operations, especially by a computer.) is a good way to describe showing a neural-network 100,000 picture of coke can and 100,000 of non-coke cans and then asking it if there a coke can in a video and geting and answer without knowing how it does it ? (and can give a different answer to the exact same image)
 
Last edited:
KazeoHin said:
I mean, ML LLMs are primarily being used for Social Media posting provocative or targeted AI generated images, Bots are taking over most of the online world to the point that I wouldn't be surprised if in 10 years time people with any amount of agency in their life mostly move away from social or internet media entirely, so yeah, AI will be so convincing at targeting ragebait or walmart/shein ads to your mom or dad it'll have no real-world use as everyone just avoids avenues where it exists, becoming functionally useless.
Click to expand...
We've been using AI to write articles and the recent AI changes have definitely made some people lose their jobs. AI is certainly going to do damage to artists who could be replaced by a 15 year old who knows how to use Adobe FireFly. I'm having so much fun with Dream Machine. The voice acting stuff is wild. Beyond these things I'm struggling to see what jobs AI could displace. It'll certainly make some jobs more efficient, which will result is people losing jobs but that may take a while.
 
Last edited:
DukenukemX said:
I'm struggling to see what jobs AI could displace. It'll certainly make some jobs more efficient, which will result is people losing jobs but that may take a while.
Click to expand...
First line customer service will be one of the first (chats or talk), like always take those claim with a grain of salt: https://www.klarna.com/internationa...of-customer-service-chats-in-its-first-month/, the 24/24, 365/365 no wait line for customer, every language and accent available, will probably be too good to pass on and with a little bit of budget the voice quality-sentence will be hard to know if it is an human or not.

Will it mean less customer service jobs or humans having more time to spent with customer with issues more complex than the simplest common situation the AI can handle and now customer using it more than they can actually access humans that have time for them, could be a mix of both.

One that displace could be the good name for it, company that sell easy programming (like UI), already started, anything of the type data entry clerk, telemarketer-sellers.

On a mid horizon cashier, drivers, lot of paralegal work (finding relevant precedent, creating giant amount of documentation to overwhelm the opposition, going through the opposition data that try to overwhelm you kind of war low job), radiologist (they are really rich and well-connected to defend their superbe profession too, so could be hard), stock trading, translator outside fancy books, movies if laws allow it. for a lot of them, there always if law* allow it (driver, translator, radiologist, oncologue etc...)

How good robots tech grow alongside (price wise as well or mostly) will determine this a lot as well, an mass produced Tesla Robot in 25 years that cost about the same as the cheaper car now and can do everything around the house and work a lot of current jobs open the door for a lot of displacement, and with the aging population that could be bring a constant immigration-work shortage-inflation crisis, people could get quite open to robot-ai worker.
 
You must log in or register to reply here.
Back
Top