ePotentia
epotentia.com
ePotentia
@epotentia.com
16 followers 180 following 30 posts
Scientific AI consultancy company talking about new developments in AI.
Posts Media Videos Starter Packs
Power-efficient models, training methods and hardware will be at least as important to the future as building more and more data centers.
The rapid proliferation of services powered by artificial intelligence, ranging from the genuinely useful to the frivolous to the unnecessary, has a common element: the need for vast amounts of electricity and water for the data centers that make them possible.
The Unsustainable Strain Of AI’s Insatiable Power Needs
This week’s Current Climate newsletter also looks at how electric cars could get much cheaper and Rivian’s RJ Scaringe on China’s EV price advantage
www.forbes.com
As AI often blindly consumes the internet, only to pollute it afterwards it's worth thinking about the future of information sharing. Do we take action to safeguard the internet? Do we trust and tolerate a limited number of LLM companies to become the source of truth for all information?
One of the reasons we chose AI for characterization as one of our main applications is its ability to accelerate research in many different fields. Glad to see it also help cancer research. :)
A new technique combines lasers, fluorescent dyes and artificial intelligence to "tag" cells with abnormal chromosomes that may become cancerous.

Read the OPN story for more! bit.ly/4os3Jpa 💡⚛️🧪
The question is whether disclaimers and basic guardrails stop the user from getting this information or just makes it less predictable.
Artificial intelligence for sound is often forgotten, but besides faking voices and making music it offers amazing potential in understanding animal behavior.
Biodiversity project takes flight 🪁

The “Ambient” initiative is using artificial intelligence to identify birdsong on the CERN sites and monitor biodiversity.

Find out more: home.cern/news/news/ce...
The economy of scale gives the big companies an edge, still they should be careful assigning fixed prices to varying usage of products in an evolving market. :)
Until the beast can escape into our phones or through robotics the immense energy and hardware needs of giant AI should make it difficult to get out of control. And while we do now have agents, ones based on our best models are still rather slow. For now let's worry about humans abusing AI. :)
Geographical bias in LLMs is a big problem. Initiatives like the EUs GenAI4Africa try to level the playing field but even when funding models, sufficient compute infrastructure is also not available both to train and deploy models. What do you think are the steps to equality?
"Artificial intelligence (AI) tools like ChatGPT, DeepSeek, Siri or Google Assistant are developed by the global north and trained in English, Chinese or European languages. In comparison, African languages are largely missing from the internet."

theconversation.com/african-lang...
African languages for AI: the project that’s gathering a huge new dataset
When a language isn’t in the data, its speakers aren’t in the product – and AI cannot be safe, useful, or fair for them.
theconversation.com
Very important initiative, but we do worry a bit about it's feasibility. Watermarks can often be filtered out by running an open-source model afterwards, meaning you can likely control and punish legal entities, but it may be different to fight malicious actors.
It's beautiful :) though in the end it shows that the molecules just differ by limited functional groups, which may limit the search space we can cover. AI can help us speed up simulations that help us explore entirely new molecular families which we may never find ourselves. :)
Realistic LLM security will be essential for all applications. Even subtle influences on peoples mental state may in fact be much more dangerous than telling people how to make nuclear weapons...
A resilient and sustainable European market is essential for our future even without the threat of war. Happy to support the European steel industry through our Horizon Europe AI research. :)
We must all demand quality instead of quantity in... everything.
The new age of AI will require a paradigm shift in peer review if it is to survive .
Thanks! So far it looks like it'll be a pleasant stay! :)
While being a large proponent of AI, we do wonder somehow about the immense lack of caring even among big companies when pushing likely cheap small llms with weird opinions in production. :) efficient yet robust AI should be encouraged
Holy shit, that sounds pretty interesting, where do we sign up? :)
Computers allowed us to run simulations on top of doing experiments and solving equations on paper. Many simulations are hitting their limits, as are classical analysis methods. Is it necessarily that bad to teach responsible AI usage in research? :)
In the end you can't lose with investing in compute. If someone suddenly invents highly efficient neuromorphic chips however then a few bubbles could pop if production could keep up. :)
Developing foundation models is a battle of extremes, the rapid progress of open-source models vs massive amounts of lost revenue that would eventually lead to the copyright owners to become obsolete
First we need legal text to be logical and understandable to serve as ethical training data. :)
There is still time to develop Talkie Toaster from Red Dwarf too :)
Hey, we encourage our AI to make the best of puns. :) More seriously though, while AI is overhyped it can also bridge a lot of the tedium in science and let you focus more on building conceptual understanding. Throwing the funding agency a bone with some bayesian optimization may be worth it. :)
It's definitely good to remember there is more than one language for our training data, especially ones used by a significant portion of the world :) of course this is also something to remember for the Chinese and it remains interesting to compare models with different biases.
Luckily on both sides exponential ai growth is constrained by producing hardware or scrap would rain from the sky forever :)