overview for Lugh

Hollywood's love of dystopian sci-fi has a lot to answer for, as it has shaped many people's ideas about the future very negatively. One of the most persistent of those ideas is that robots will only be owned by the 1%, who will use them to subjugate everyone else.

Reality is shaping up to be different. Free, open-source AI is the equal of anything privately controlled. Robotics too looks like it is following a similar trajectory. The Berkeley Humanoid Lite is built with off-the-shelf and 3D-printed components and costs just $5,000.

Contrary to doomerist fantasies, with decentralized renewable energy, and open-source AI & robotics - it seems hard to believe the 1% will own everything in the future.

13

Is Open-Source vs. Proprietary AI the real AI race, not US vs. Chinese AI, and is Open-Source winning? (self.futurology)

submitted 2 months ago by Lugh to c/futurology

1 comments fedilink

One of the distortions of AI commentary is that so much of its focus is on Venture Capitalism. Because many people are incentivized to talk about where the big money is flowing, they ignore outside their bubble. Meanwhile, often the really significant things happen elsewhere.

With AI that 'really significant' thing - is that free open-source AI is the global future, far more than the VC darlings like OpenAI. Not that the people pouring hundreds of billions of dollars into the likes of OpenAI are likely to admit that.

There are more signs of this as recently as this week. Yet again, free open-source AI (in this the Qwen3 family from Alibaba) is not only equalling the best of the investor-funded AI, they are bettering it in some metrics.

The VC's thinking is that one of their bets will make big & generate trillions in revenue, but it seems hard to believe when all over the world people can pick up what you're trying to sell for free.

19

Waymo's latest research shows its self-driving cars have 80-90% fewer accidents than human drivers, and in future could possibly save 34,000 U.S. lives annually if they replaced all human-driven cars. (self.avs)

submitted 2 months ago by Lugh to c/avs

5 comments fedilink

Waymo's peer-reviewed study in Traffic Injury Prevention, PDF, 58 pages found its self-driving cars safely drove 56.7 million miles across four U.S. cities without a human safety driver. With 80-90% level reduction for different types of accidents.

56.7 million miles is a tiny fraction of the overall US miles driven, only about 0.002%. Current self-driving AI wouldn't be as good for all road types and conditions. But it will get there, the only question is when. When it does that 80-90% reduction in accidents means 34,000 lives saved in the US, and hundreds of thousands globally - every single year.

The day is going to come where the public conversation is going to be about banning human driving, like no-seatbelts and indoor smoking before it. I've a suspicion the same people who said losing a few hundred thousand lives to 'herd immunity' will be telling us that those 34,000 dead a year are a price worth paying, so they don't have to change anything about their lives or routines.

-4

A new crowd-trained way to develop LLMs over the internet could shake up the AI industry with a giant 100 billion-parameter model later this year. (archive.ph)

submitted 2 months ago by Lugh to c/futurology

1 comments fedilink

25

The AI jobs crisis is here, now: It's not coming, it has already arrived. (www.bloodinthemachine.com)

submitted 2 months ago by Lugh to c/futurology

1 comments fedilink

24

A Swedish company deploying underwater tidal kites in the Faroe Islands, says 500 of them would supply 100% of Alaska's electricity needs. (www.emergingtechbrew.com)

submitted 2 months ago by Lugh to c/futurology

6 comments fedilink

India says it has achieved key milestones for the 2028 launch of an Indian Space Station to join the ISS and Tiangong Space Station. in c/futurology

[–] Lugh 0 points 3 months ago

So if all goes to plan there should be three space stations in 2028 - when will there be a fourth? Russia has plans for an Russian Orbital Service Station. While no one doubts Russia has the necessary technical expertise, can it spare diverting the cash from its war in Ukraine?

Presumably that will be a commercial one someday, it doesn't seem like any other countries are in the pipeline for developing one. The ISS is slated to be deorbited around 2030. The US & ESA/Canada might well part ways when the ISS is decommissioned, and it seems unclear what will replace the ISS, and exactly how much longer it will last.

China's new battery material charges in secs, can transform EV industry in c/futurology

[–] Lugh 4 points 3 months ago

In fairness to China, they are now the world leader in battery tech, and delivering on their promises.

In a world first, Chinese scientists have demonstrated a brain-spine interface that enables paraplegic patients with severed spinal cords to walk again. in c/futurology

[–] Lugh 3 points 3 months ago

Yes, there's also the question of how long it lasts; but still, an amazing achievement.

In a world first, Chinese scientists have demonstrated a brain-spine interface that enables paraplegic patients with severed spinal cords to walk again. in c/futurology

[–] Lugh 8 points 3 months ago

This is only in 4 patients so far yet the results look amazing. 20 million people globally are living with some form of spinal cord injury. Hopefully insights gained from this work will quickly mean treatments for what was once seen as incurable.

The crypto-industry is spending hundreds of millions to push its agenda in Washington, claiming the U.S. is in decline and advocating for global "network states" run by business owners. in c/futurology

[–] Lugh 1 points 3 months ago (1 children)

I always thought network states were just online communities

Perhaps for some, but the idea of making them physical territories has always been there too, and arguably is the idea in the ascendant now.

Dark Energy experiment shakes Einstein's theory of Universe in c/futurology

[–] Lugh 2 points 3 months ago

From Galileo to Newton to Einstein, we've so many times had our ideas of the Universe completely and utterly upended. It seems to me when people talk about dark matter and dark energy, they're just tinkering with current ideas. When we really find the explanation, our ideas of what reality is will be completely shaken once again.

Mars could have an ocean's worth of water beneath its surface, seismic data suggest in c/futurology

[–] Lugh 4 points 3 months ago

Rock From Mars has reached Earth so many times via asteroid ejecta, I assume the opposite has happened. In which case if there is life there surely the probability is it comes from Earth?

Baidu’s new multimodal foundation model, ERNIE 4.5, outperforms OpenAI’s GPT-4.5 on several benchmarks—while being 99% cheaper. in c/futurology

[–] Lugh 5 points 3 months ago (1 children)

I wonder how long before the AI investor bubble bursts? OpenAI are still talking about raising hundreds of billions of dollars.

Outperformed by Chinese Open-Source AI, US firms want their government to ban it. in c/futurology

[–] Lugh 7 points 3 months ago

Exactly, At best, it's a stay of execution.

Scientists Just Discovered an RNA That Repairs DNA Damage – And It’s a Game-Changer in c/futurology

[–] Lugh 6 points 4 months ago (4 children)

According to the scientists, knowledge about the role of NEAT1 methylation in the recognition and repair of DNA damage could open up new therapeutic options for tumors with high NEAT1 expression. However, it must first be clarified whether these results, which were obtained in simple cell systems, can also be transferred to complex tumor models.

I wonder is AI developments can speed up this process of going from lab to (hopefully) a therapeutic treatment that is available.

Chinese AI Video Generators Unleash a Flood of New Nonconsensual Porn in c/futurology

[–] Lugh 13 points 4 months ago (2 children)

How are we supposed to police this? A vast global Orwellian monitoring system seems even worse.

Pentagon to give AI agents a role in planning, operations in c/futurology

[–] Lugh 2 points 4 months ago

Indeed, Israel, for one, already uses Made-in-America AI models in war.

Indiscriminately murdering dozens of children a day has been a real advertisement of how brilliant these are.