John Carmack foresees a breakthrough in artificial general intelligence by 2030

midian182

Posts: 9,745   +121
Staff member
Forward-looking: While AI has been at the forefront of most tech industry conversations this year, the new wave of generative AI is still far off the concept of an artificial general intelligence (AGI). However, legendary developer John Carmack believes such a technology will be shown off sometime around 2030.

Carmack, of course, is best known as the co-founder of id Software and lead programmer of Wolfenstein 3D, Doom, and Quake. He left Oculus in December last year to focus on Keen Technologies, his new AGI startup.

In an announcement video (via The Reg) revealing that Keen has hired Richard Sutton, chief scientific advisor at the Alberta Machine Intelligence Institute, Carmack said the new hire was ideally positioned to work on AGI.

It's often pointed out that the term artificial intelligence is a misnomer when used to describe the likes of ChatGPT. The large language model-powered tools of today don't actually 'think' like a person, even though they can appear that way. An AGI, on the other hand, acts just like a human, able to learn and reason and accomplish any task it is given. Data from Star Trek and, more worryingly, HAL 9000 from 2001: A Space Odyssey are often cited as examples of advanced AGIs from science fiction.

Scientists and researchers argue that a true AGI could be decades away. Sutton says that current coding techniques make creating an AGI infeasible, adding that 2030 is a target for a "prototype AI to show signs of life."

In August last year, Carmack announced that he had raised $20 million in initial funding for Keen. That's a tiny amount compared to other AI companies – Microsoft has poured more than $13 billion into OpenAI – but Sutton and Carmack believe they can have a huge impact on the industry.

"Nobody has line of sight on a solution to this today, we feel there is not that much left to do," Carmack said. "There are fundamental research questions that need to be answered, and we have internal projects and angles of attack," he added. The creator of the FPS genre said that he expects future textbooks on how to build AGIs will include a chapter on Keen and its upcoming work.

"We are six, seven, eight years out from something really big and important being publicly visible."

Carmack later suggested that the state of AI today can be traced back to his own work on Quake as it was the game that really sparked the demand for GPUs, though he also credits Nvidia CEO Jensen Huang with recognizing the potential for GPUs to handle other computing tasks.

Back in July, OpenAI said it was aware of the potential dangers of creating an AI smarter than people, including it causing the extinction of the human race, and was putting together a team to mitigate the risks.

Permalink to story.

 
Along with breakthroughs in batteries, fusion and flying cars.

The same speech is brushed off every year, like a Mad-Libs, with new nouns inserted.

Talk is cheap, demonstrate something that is functional, affordable, and useful.
 
Along with breakthroughs in batteries, fusion and flying cars.

The same speech is brushed off every year, like a Mad-Libs, with new nouns inserted.

Talk is cheap, demonstrate something that is functional, affordable, and useful.
As I see it, AI, in the context of search engines that have employed it, is anything but useful and is more akin to a glorified search engine.since its answers, though they might sound eloquent and authoritative, have to be verified for accuracy. I just don't see the point of AI in search engines ATM. In other areas like medicine, however, it sounds like it is useful. In those areas, you can bet that its responses are extensively scrutinized for accuracy - so there, too, it still has to have humans looking over its shoulder to verify its accuracy. To me, this implies that AI cannot presently be trusted to give good results.

IMO, I think Carmack has spent too much time in his fantasy realms to be commenting on life in the real world.
 
Maybe by 2050 we'll have something approaching actual AI. What we have now is special needs LLMs that pretend like they're anything but a janky search engine.
 
2 years ago the idea of AI was still really only in ScFi films. These days a use the current gen of AI for graphics, to help with coding tasks and to suggest future projects - it's suggestions are fairly decent. Considering the current acceleration of AI, It doesn't seem like a fantasy to think we'll have general AI by the end of the decade.
 
2 years ago the idea of AI was still really only in ScFi films. These days a use the current gen of AI for graphics, to help with coding tasks and to suggest future projects - it's suggestions are fairly decent. Considering the current acceleration of AI, It doesn't seem like a fantasy to think we'll have general AI by the end of the decade.
And using more energy than Crypto mining.
 
As long as they keep it out of government. Intelligence & government should NEVER be used in
the same sentence.
 
And using more energy than Crypto mining.
Interesting point. t appears that most of the energy is consumed during the training process. ChatGPT uses about 10GWh to fully train (equivalent to 1000 homes for a year) but only 0.0002Kwh to answer a query. Obviously there are a lot of queries per day but I found it difficult finding an exact figure.

The energy usage for crypto was considerably more though at 127TWh per year. 1TWh = 1000000GWh. Google used about 15TWh of electricity. Aluminium smelting consumes about 40,000GWh per year . It's also debatable whether anything positive came from crypto unless you were a South American drug lord wanting to launder money.

It might make an interesting article covering energy usage by different industries. Also a chart of the energy usage for an average family in different countries. Maybe go wild and show the energy usage in different forms of transport.
 
Back