Five Things AI: Blackbox, 2024 Campaign, ScarJo, Theory of Mind, Human
Let's wrap this week in AI!
Hi and welcome back to Five Things AI! My name is Nico Lumma and every week I like to share five interesting articles about AI with you. So instead of researching for hours on the web, just read these five articles here and you are all set. As I’m saving you a couple of hours every week, I’d really appreciate it if you could go paid and spend the equivalent of one large café latte on this newsletter.
The big news from Germany this week: DeepL raises $300 mio at a $2 billion valuation for their AI based translation platform. This is an amazing achievement and I sure hope that they will grow their offering accordingly. After seeing the GPT-4o demo last week, I think the race is on to get embedded translation with a little thingy in your ear, soon.
Would that be awesome? Or just Babylon, again? And what would that mean to my Duolingo Learning Hebrew Streak which is now at 170 days?
I think we need to focus more on the cool things that we can achieve with AI and not on all the challenges and shortcomings of AI. By the way, Humane is looking for a buyer. Humane solved a problem not many people had. This is the way innovation works. Now we know better and some investors lost some money to find this out. The next investments will build upon this experience.
So, please read these Five Things AI and if you only have time for one article, read the first one. The findings at Anthropic are really a game changer for our understanding of LLM.
AI Is a Black Box. Anthropic Figured Out a Way to Look Inside
“The company’s researchers have identified the combination of artificial neurons that signify features as disparate as burritos, semicolons in programming code, and—very much to the larger goal of the research—deadly biological weapons. Work like this has potentially huge implications for AI safety: If you can figure out where danger lurks inside an LLM, you are presumably better equipped to stop it.” - this is really fascinating stuff that has the potential to give all of us a better understanding of how AI works and what it is capable of.
A.I. Promised to Upend the 2024 Campaign. It Hasn’t Yet.
I wouldn’t be so sure about it. There’s still plenty of time until the election to push AI generated content on a massive scale. And the later this gets starter, the more problematic it will be to stop it.
ScarJo-OpenAI fight focuses the AI safety debate
Don’t mess with Scarlett Johansson. OpenAI just opened a huge can of worms.
AI Outperforms Humans in Theory of Mind Tests
Does AI really get us? After we now have a better understanding of how the blackbox of an LLM works, we will also see more advances in this field.
The Big AI Risk Not Enough People Are Seeing
“The new AI products coming to market are gate-crashing spheres of activity that were previously the sole province of human beings. Responding to these often disturbing developments requires a principled way of disentangling uses of AI that are legitimately beneficial and prosocial from those that threaten to atrophy our life skills and independence. And that requires us to have a clear idea of what makes human beings human in the first place.” - just think about that for a minute.
If you missed last week’s edition of Five Things AI, you can read it here:
:extract_focal()/https%3A%2F%2Fmedia.wired.com%2Fphotos%2F6644e7c11055f1e066c3b4ce%2F191%3A100%2Fw_1280%2Cc_limit%2FAI-Chatbots-Fast-Forward-Newsletter-Business-GettyImages-2022303132.jpg)
That’s it for Five Things AI this week! 🤖
— Nico