I think the major advancements are outside the textbox in this last year: video generation, robotic models like Helix, world models, Genie 3.
Even for text, Deepseek R1 was this year, and agentic and coding AI has made progress on length/complexity of tasks. The rise of MoE architecture in the open/local model space has made it possible to run useful models locally on hardware under like $2K, something I didn’t expect for a long time.
Having it hallucinate a valid url that is spoofing the site I’m looking for feels less likely than someone managing to game SEO. Eclipse is a good example: the first result in Google is eclipseide.org, not eclipse.org.
The video specifies that the drug is infused over 8-10 hours. Probe placement - again, as depicted in the video, because I don’t see a real methods section - should take about 1-2 hours. The video isn’t clear if this is interactive MRI or just a preop scan that is then loaded into a stereotactic navigation system in a regular operating room, but the former would add another hour at least. MRI is not fast.
Does ASML's investment portend a pivot to specialized, on-prem, enterprise models? No need to be the frontier general knowledge or even coding model, but instead an EU-based AI creator for things like chip design, pharma, automotive, etc?
Not even just for on-premise deployments, even for cloud settings. Google has demonstrated that you can profit very much from having your own specialized AI chips to bring down cloud costs. Maybe the EU with all the talks about giga AI factories is also planning to go in that direction instead of continuing to rely on overpriced NVIDIA chips.
Given current leaderships; it’s not hard to imagine scenarios where access to leading AI models from the US or China could be cut-off, restricted or otherwise compromised.
ASML, while European, has significant exposure to Taiwan’s semiconductor industry and is therefore vulnerable to risks from both sides. At the same time, the EU is aware of the danger of falling behind in its AI capabilities compared to the US and China.
In that light, the investment seems likely to be a mix of tax efficiency, building goodwill with the EU leaders, and a strategic hedge by ASML to ensure some degree of AI capability closer to home.
While I associate Mistral with LLMs, the electric design automation software used for planning and designing chips already uses machine learning/reinforcement learning for some approaches. AI could play an even greater role in chip design in the future.
Llms are fundamentally different algorithms and problem space to IC design and production. Why would mistral be helpful?
I dont see how even the algorithms involved translate well. IC design is closer to a physics simulator connected to a heuristic optimizer. Mabie some ideas from alfageometry or alfafold could be applied, but thats not the kind of research mistral is doing.
And there are big players with existing expertise in the IC design space. Why not just fund them to do more research?
Even for text, Deepseek R1 was this year, and agentic and coding AI has made progress on length/complexity of tasks. The rise of MoE architecture in the open/local model space has made it possible to run useful models locally on hardware under like $2K, something I didn’t expect for a long time.