Great idea. Submit Your Toughest Questions for Humanity's Last Exam.

Center for AI Safety

Submit Your Toughest Questions for Humanity's Last Exam | CAIS

CAIS and Scale AI are excited to announce the launch of Humanity's Last Exam, a project aimed at measuring how close we are to achieving expert-lev...

#ai #benchmarks

Of je nu sceptisch of enthousiast bent over AI, het is een gegeven dat Microsoft en andere bedrijven het volop gaan integreren in hun toepassingen waardoor we er nauwelijks meer aan kunnen ontkomen, #microsoft #office365 #AI #copilot

Linkedin gaat haar data gebruiken voor het trainen van AI, nu kun je een opt-out aangeven en wordt EU data kennelijk nog niet meegenomen. Maar is die scheiding zo scherp te trekken als het gaat om data zoals in Linkedin (netwerk met veel relaties) ? #AI #microsoft #linkedin #privacy

Er zitten veel aannames en verwachtingen in onderstaande tekst uit de troonrede: "Om te vermijden dat het personeelstekort in de zorg onbeheersbaar wordt, is het urgent de administratietijd te halveren, zodat er meer tijd overblijft voor de patiënten. Dat vraagt om innovaties in de sfeer van digitalisering en slim gebruik van kunstmatige intelligentie." #politiek #troonrede #AI #gezondheidszorg

Het zou prettig zijn als de vele verwachtingen die de regering heeft mbt AI ook zouden worden gecombineerd met een omvang van investeringen die daarbij hoort. #AI #regeerprogramma #schoof #politiek

Playing around with GPT-4o1 preview, it's really good in science, math, chemistry etc., an example.

ChatGPT

ChatGPT - Work Calculation for Reaction

Shared via ChatGPT

#AI #chemistry #thermodynamics #science #GPT4 #GPT401

GPT-4o1 is niet een nieuw model zoals de naam ook al aangeeft maar een verbetering van GPT-4o , ook heeft GPT-4o1 geen toegang tot internet , kun je geen documenten uploaden om mee te nemen in de analyse. Dat gezegd hebbende, het is indrukwekkend welke prestatieverbetering ze, met het bestaande model, hebben weten te realiseren. Met name voor specifieke taken waarbij redeneren van belang is.

Introducing OpenAI o1

Introducing OpenAI o1

#AI #openAI #GPT401

I tried this AI generated podcast feature on NotebookLM and i am very impressed, it even made a rational sounding podcast from a document in dutch.... I am afraid we will be swamped by AI generated podcasts....

Google

NotebookLM now lets you listen to a conversation about your sources

NotebookLM is releasing Audio Overviews, which turns your sources into an engaging discussion.

#AI #Google #Podcast #NotebookLM

AI is about compute, data and algorithms. This paper looks at the global compute available in the public cloud. Interesting paper. "Compute North vs. Compute South: The Uneven Possibilities of Compute-based AI Governance Around the Globe"

OSF

#AI #compute #geopolitics #cloud

Important observation to think about by @npub143w0...pykj 🤔

IOC.exchange

Matthew Green (@matthew_d_green@ioc.exchange)

One of the things we need to discuss is that LLMs listening to your conversations and phone calls, reading your texts and emails — this is all go...

#ICT #AI #privacy #LLM #security

Even if all AI models wouldn't improve one bit from today, the impact in various applications on smartphones, education, business etc. will still be enormous. The push for implementation in more products the coming months will also be big because of the sometimes ridiculous amounts of money invested. So we are in for a rough ride. Forget about "the bubble bursting", thinks will get hectic (in my view). And ofcourse big AI companies will keep overpromising and not delivering 😀 #AI

Lot of excitement around Reflection 70B , a new model that outperforms many existing models apparently, it tries to correct hallucinations

MarkTechPost

Reflection 70B: A Ground Breaking Open-Source LLM, Trained with a New Technique called Reflection-Tuning that Teaches a LLM to Detect Mistakes in Its Reasoning and Correct Course

Reflection 70B: A Ground Breaking Open-Source LLM, Trained with a New Technique called Reflection-Tuning that Teaches a LLM to Detect Mistakes in I...

#ai #reflection70b

Promising preliminary findings with regard to AI tutors in education. Matches my own experience with ChatGPT which is quite useful for explaining difficult subjects.

Harvard Gazette

Professor tailored AI tutor to physics course. Engagement doubled. — Harvard Gazette

Preliminary findings inspire other large Harvard classes to test approach this fall

#education #harvard #ai #tutor

A free book in PDF, "Person, Thing, Robot: A Moral and Legal Ontology for the 21st Century and Beyond." https://direct.mit.edu/books/oa-monograph/5641/Person-Thing-RobotA-Moral-and-Legal-Ontology-for #MIT #Robot #AI #moral #legal #free #book

It's not enforceable , there are no sanctions, still it's a good thing some common norms and principles are agreed upon in this framework. Fully in line with the AI Act by the way.

Shaping Europe’s digital future

Commission signed the Council of Europe Framework Convention on Artificial Intelligence and human rights, democracy and the rule of law

The Commission signed the Council of Europe Framework Convention on Artificial Intelligence on behalf of European Union.

#AI #EU #europe

DisTrO sounds exciting, A new way to train large AI models that drastically reduces the need for super-fast internet connection, potentially allowing training on regular home computers and internet connections. The promise is that DisTrO cuts down the amount of data that needs to be shared between computers during pre-training and during fine-tuning with several magnitudes. But only a preliminary report , not a reviewed paper yet.

GitHub

GitHub - NousResearch/DisTrO: Distributed Training Over-The-Internet

Distributed Training Over-The-Internet. Contribute to NousResearch/DisTrO development by creating an account on GitHub.

#AI #training #distributed #DisTrO

Ik zag in het OV een student op zijn laptop statistiek opgaven maken met ChatGPT als hulpje, benieuwd hoe vaak dat inmiddels gebeurt (veel denk ik) , los van de vraag of dat nu slim is, ik denk dat het gebruik door scholieren/studenten niet moet worden onderschat. #ai #chatgpt #onderwijs

OpenAI will probably say anything to increase it's value, make it more attractive to partners, acquire funds to stay alive with their burn rate. So it sounds promising GPT-Next, 100x as powerful as GPT-4, but i believe it when i see it. #AI #GPTnext #openai #GPT4

"Council of Europe Framework Convention on Artificial Intelligence and Human Rights, Democracy and the Rule of Law" #AI #treaty https://www.coe.int/en/web/conventions/full-list?module=treaty-detail&treatynum=225

ChatGPT heeft 200 miljoen wekelijkse gebruikers en dan moet de samenwerking met Apple nog beginnen.

DPG Media Privacy Gate

#AI #OpenAI #ChatGPT