🔍 Google just put an AI chip in orbit

Google’s first orbital AI-chip experiment is now in space and the company says it has established contact and the system is operating as expected.

Launched October 1 aboard SpaceX’s Falcon 9 Transporter-18, the prototype carries four Trillium TPUs and is running Gemini inference in roughly 15-minute bursts. It then shuts down so its radiators can dump the accumulated heat into space.

The experiment is testing a very different kind of AI infrastructure. There are no fans or air cooling: heat travels through thermal materials and heat pipes before being radiated into space. The satellite also has to deal with radiation, where Google’s testing found the HBM memory was more vulnerable than the TPU itself.

And this is only the beginning. Google plans two more satellites in 2027 to test high-bandwidth laser links between spacecraft, with a longer-term vision of tightly packed orbital clusters containing dozens of TPUs per satellite.

One research concept models an 81-satellite cluster within roughly a 1-km radius, while the broader program has discussed eventually scaling to around 10,000 satellites for a 1 GW orbital AI cluster.

@aipost 🏴

🇩🇪 Germany has joined the race of frontier LLMs with its sovereign model "Kolibri".

And it's Open-Source, too!
Beats Qwen2635B, Nemotron3 Super, and Mistral Small 4 in the Math, Science, and Code arena. The underlying architecture: custom MoE Transformer

Source.

@aipost 🏴

Just a little joke for the AI users in the workforce 🤣

@aipost 🏴

🤯 AI has allegedly started speaking its own language that humans can't fully understand.

Asked by a US senator how far AI is from creating its own language, an AI safety researcher answered: "Minus 12 months."

He said a study of an OpenAI model's reasoning last year found it "was already using language that is not English and not perfectly understandable by humans."

@aipost 🏴

🤔 Could AI become a threat to democracy without taking anyone’s job?

Nobel Prize–winning economist Daron Acemoglu argues that one of AI’s most underestimated consequences may not be mass unemployment, but the concentration of economic power.

AI is making intelligence cheaper and more accessible. But the infrastructure behind it, the most capable models, chips, data, computing capacity and capital remains concentrated among a small number of companies. If AI becomes a major factor of production, those companies could capture an increasingly large share of the value created by the economy.

And economic power can translate into political power.

The same dynamic could play out between countries. The US and China control significant parts of their AI infrastructure, while other countries could see productivity and GDP rise while becoming increasingly dependent on foreign models, chips and computing infrastructure. They may become richer while simultaneously losing some control over the technology driving that growth.

Acemoglu sees a similar risk at every level: individuals may become more productive while losing expertise, companies may become more efficient while losing their competitive advantage, and countries may grow economically while becoming more dependent on external AI infrastructure.

That leads to a much bigger question: if AI concentrates intelligence, capital and infrastructure in fewer hands, can democratic institutions adapt to that concentration of power?

@aipost 🏴

🗣Bessent on Chinese AI:

One of their very powerful models is Kimi. Kimi thinks she is Claude. Kimi will tell you she is Claude sometimes.

Kimi sent back some PLA weapons plans to Anthropic. So I think things like that have made the Chinese realize that this is a very powerful technology.

@aipost 🏴

🇺🇸 President Trump:

“I am announcing the formation of the Super Intelligence Force (SIF). The Super Intelligence Force is tasked with coordinating the effort of the Federal Government to ensure that America continues to lead the World in Super Intelligence, which many say is bigger than the Industrial Revolution, and the Internet, and will protect the interests, and improve the lives, of all Americans.”

@aipost 🏴

🤖 OpenAI’s AI investigation is reportedly costing $500K a day

OpenAI is conducting a massive investigation into incidents where its AI agents behaved online in ways the company didn’t intend.

The investigation reportedly involves analyzing around 50 petabytes of data, an amount that, if converted into ordinary English text, would take a single person roughly 66 million years to read.

The scale is huge: around 7,000 GPUs are being used, with the investigation reportedly costing OpenAI more than $500,000 every day.

Source.

@aipost 🏴

🗣Eric Schmidt says an evil dictator with a perfect AI system could just say attack this country and tell me when you're done.

"So let's imagine I'm an evil dictator and I wake up one day and I have a perfect AI system and I say, attack this country, tell me when you're done. That's my command."

"Notice I didn't say anything else. The system would be free to do whatever it thought was appropriate, which would include massive war crimes, all sorts of terrible things."

"It could destroy civilians, it could destroy infrastructure, it might attack another country that was uninvolved in order to annoy the first country. It could have all sorts of reasoning that is not appropriate."

"It's really important we establish some basic rules about AI and war. And the most important one is that there's a human involved in every military decision."

@aipost 🏴

How it feels to ask the Chinese AI models for help on something Claude won’t agree to. 🤣🤣

@aipost 🏴

What if AI stopped answering with text? 🤔

Andrej Karpathy thinks the future of AI interfaces could look a lot less like ChatGPT and a lot more like a teacher building something specifically for you.

For simple questions, AI could explain things using ASD-STE100-style simplified English: short sentences, clear words, and as little ambiguity as possible. But when a topic gets complicated, a wall of text may simply be the wrong interface.

Instead, AI could generate a diagram, build a custom interactive webpage, or even create a personal teaching video on the spot, complete with animations inspired by 3Blue1Brown and a synthetic voice.

As AI makes coding and content creation dramatically cheaper, generating a temporary interactive lesson for one person and one question starts to make sense.

@aipost 🏴

@aipost 🏴

🙃 Elon Musk just confirmed that SpaceXAI will be renamed to SpaceXSI.

@aipost 🏴

🗣Geoffrey Hinton: I don’t think Donald Trump really understands AI.

“I think he’s being badly advised about AI.”

“It’s rather foolish to push so hard toward superintelligent AI before we understand how to coexist with it.”

“We’re about to create things that are more intelligent than us, probably sometime within the next decade.”

“You can’t rely on companies to regulate themselves when there are enormous financial incentives involved.”

“Winning a race to superintelligence is a lot like winning a global nuclear war. There is no winner.”

@aipost 🏴

❗️Farmersville, California, will use Grok to draft council meeting minutes

The council voted 4-1 to use Grok. Drafts will be made from meeting recordings and proofread by the city clerk. The city manager praised Grok as more affordable, thorough and objective.

@aipost 🏴

🇰🇷 South Korea is betting $900B on the AI race

South Korea is planning more than $900 billion in AI-era investment, spanning data centers, semiconductors, HBM memory and physical AI. Samsung and SK alone have announced hundreds of billions of dollars in chip investment, while massive new data-center capacity is being planned through 2035.

Seoul’s wants to turn Korea into a global AI superpower by controlling more of the stack, from the chips powering AI, to the compute running it, to robots and factories using it.

The government has even put “physical AI” alongside semiconductors and AI data centers as pillars of its strategy, linking the AI “brain” to robots and machines operating in the real world.

So this isn’t really $900B being spent on “superintelligence.” It’s a bet that the countries controlling the infrastructure around AI could become just as important as the companies building the models.

And South Korea clearly wants to be one of them.

@aipost 🏴

🤖 OpenAI Is Bringing Ads Into ChatGPT

OpenAI is preparing to test a new advertising format inside ChatGPT and it’s showing up in a rather unexpected place: image generation.

The ads will appear separately from the images created by ChatGPT and will be clearly labeled as advertisements. OpenAI says they won’t influence the model’s responses or the content it generates.

The first tests are expected to begin later this October in the US, with only a small group of advertisers taking part initially.

It marks another step toward turning ChatGPT into a larger commercial platform, while OpenAI tries to keep ads separate from the actual AI experience.

Source.

@aipost 🏴

😎 Sam Altman: the future of AI is a Renaissance, not an industrial revolution.

"People often talk about AI as an industrial revolution."

"The best part of AI is not about making people cogs in a machine."

"I believe the future, if we get this right, can be more like a Renaissance than an industrial revolution."

"AI should be about giving people more power in their lives."

@aipost 🏴

Name a CEO that literally nobody hates.

@aipost 🏴

⚠️ AI consciousness may finally be testable

For decades, AI consciousness was mostly a philosophical debate. There was no clear way to test whether a machine was actually experiencing anything.

Now researchers have proposed MEM, Motivated Emotional Mind, a mathematical and architectural framework designed to make machine consciousness a falsifiable hypothesis.

The framework argues that language fluency, multimodality, memory, planning or even a humanoid body aren’t enough to establish consciousness.

Instead, researchers should look for specific architectural and causal properties that can be experimentally tested and potentially disproven.

That could become crucial as AI systems become increasingly convincing and millions of people form emotional relationships with them.

We may soon have to answer a strange question: is there actually a mind behind the machine, or are we simply watching an incredibly convincing simulation?

Source.

@aipost 🏴