🧙🏼 o1 has an IQ of ≈120

This issue is brought to you by

Companies will pay to meet your most talented friends

Share tech job opportunities with your network and earn big rewards. Join Draftboard now and start receiving up to $10k for every successful referral!

Learn More

Happy Monday, wizards.

Did you already try OpenAI’s o1 model and faced this problem too? 😄

Dario’s Picks

The most important news stories in AI this week

1. o1 scores 120 on Mensa’s IQ test, a cut above all other models. Maxim Lott has been tracking AI models’ performance on the IQ test made by Mensa Norway – o1 scored 120, well above the average human IQ.

    

   To eliminate the possibility of o1 already being trained on the Mensa test, giving it an unfair advantage over older models, the author did a secondary test for all models using an “offline” test version as well. In this test, all models performed worse but the lead of OpenAI’s new o1 model to the second best performing model (Claude 3.5 Opus) was still really big.

• ‎ Why it matters‎ ‎ I’d take the score of 120 in IQ with a grain of salt – the “offline” version of the test yielded a lower IQ score for all models, but the differences between the models is still pretty consistent. Also, remember that these tests don’t reflect the full spectrum of human intelligence at all, but focus on a narrow set of logic and reasoning tasks.

   

  PS take a look at the article’s first example of the most difficult question o1 was able to solve. I tried it myself and failed! AI is not only more knowledgeable than most humans, but also excels in pattern recognition tasks like this. No wonder all the CAPTCHAs around the web are waay more difficult now.

Continued after the ad… ‎

From our partners

Doing the same boring work again and again is exhausting.

What if you had a personal AI assistant who could do the job for you?

Register for this free 3 hour AI Mini Course & learn how to build it in just 3 hours

2. Google’s Audio Overview feature turns information into an engaging, AI generated discussion. Prepare to have your mind blown – the Audio Overviews feature takes documents, slides, charts or simply a URL, and turns it into an engaging audio discussion between two AI hosts. It sounds close to a natural discussion between two experienced radio hosts. For context, this new feature is part of NotebookLM, a research assistant developed by Google and powered by Gemini 1.5 Pro.

    

   I’ve made simple demonstration for you below. This was generated using one click(!) and giving the URL of the article mentioned above about OpenAI’s o1 model getting 120 in IQ score.

• ‎ Why it matters‎ ‎ The realism of this is astonishing. Audio Overviews is capable of turning the boring information we sometime have to consume into something much more engaging – with zero effort. I can honestly count the number of wow-experiences with AI like this on one hand.

3. Memo is changing the way people study. A while back, we reviewed the most popular tools that can turn study materials (like a PDF deck) into notes, flashcards and quizzes in seconds. Memo has sailed up in popularity lately, and reportedly has a user base of over 100,000 students now. The backstory: Memo—formerly known as PDF2Anki—was created by two medicine students who used flashcards to memorise big amounts of knowledge during their studies, and decided to streamline the manual creation of these cards with AI.

    

   ‎ Why it matters‎ ‎ I can’t help but imagine the amount of hours I would have saved if this existed while I was a student. Under the hood it uses frontier AI models to instantly create helpful study notes. If you’re a student, I’d warmly recommend checking it out.

Affiliate disclosure: To cover the cost of my email software and the time I spend writing this newsletter, I sometimes link to products and other newsletters. Please assume these are affiliate links. If you choose to subscribe to a newsletter or buy a product through any of my links then THANK YOU – it will make it possible for me to continue to do this.