The Skeptic AI Enthusiast #38

I was in AI before it became cool.

My take on the news

• Now you can call ChatGPT by phone:
  I’m dead serious here; it’s not an overdue April Fools joke. US phone users can call 1-800-CHATGPT and voice their requests to ChatGPT—not even a smartphone is required. You can even ask follow-up questions in a conversation that could last up to 15 minutes.
  I understand that current AI voice interaction makes this service possible. What puzzles me is the idea behind this initiative. OpenAI declared that “its goal is to make ChatGPT more accessible if someone does not have a smartphone or a computer handy.” That would have made sense in the nineties or early two-thousand, but now billions of people have smartphones and very few basic phones. Further, technology haters will not embrace the phone service; they don’t want to deal with technology, period.
  I will assume that the 1-800-CHATGPT service is just a publicity stunt that will not be supported in 2026.
  

• We are now in the middle of the CES 2025 show, which presents crazier than useful new products that, in many cases, look to be more intended to showcase a company than to offer real value to the user. The advantage of the CES show is that everybody gets this, and they know it’s mostly for fun than anything else.
  But in the middle of this craze, there are a few genuinely influential releases, like the Nvidia ones, that I comment on in the following items.
  

• Nvidia builds a "Personal AI Supercomputer":
  For $3,000, you can get the hardware needed to train a complete deep-learning AI system so you can ditch your reliance on cloud services. Jensen Huang, founder and CEO of Nvidia, made the announcement in a high-profile keynote.
  We know this personal AI supercomputer won’t be relevant for moving the needle of Nvidia's business (their main business will stay to provide high-performance chips to AI data centers), but to me, it’s a product that makes lots of sense: it can pay its cost just by training a single deep-learning AI system.
  

• If Nvidia’s hardware announcements weren’t enough, it also released the Cosmos framework, a collection of AI deep-learning models trained with videos of the physical world. Cosmos models can “predict and generate physics-aware videos.” Developers can, for instance, use Cosmos models directly to generate physics-based synthetic data to train their robotic simulations.
  I think that physical modeling is one area where AI models have been trailing. We all saw how Sora's videos showed some physically impossible details, such as objects floating in the air with no support or appearing or disappearing objects. This indicates that Sora doesn’t understand the physical world, and precisely, this aspect is what efforts like the Cosmos framework aim to advance.
  Last week, I discussed Google’s Veo system, which is kind of similar. This is not a coincidence; it shows the need for physical awareness to advance AI models.
  

• In its CES keynote, Panasonic announced its partnership with Anthropic to develop health-related AI extensions for its Panasonic Well division. They intend to launch family-oriented health support products like Umi, the Family Wellness Coach. Umi will include “the usual suspects” for this kind of application, like conversational AI support, a dashboard, and more.
  Panasonic (and most big Japanese companies, for that matter) have been reluctant to “contaminate” their products with crazy AI features, but perhaps now it’s mostly a matter of survival. My readers know I advocate against introducing AI in companies for the sake of AI, but at a certain point, companies fear lagging behind to a point of no return.
  I wouldn’t bet on the success of the Umi initiative, but hey, let’s at least give them the benefit of the doubt!
  

This week’s blog piece

If you are subscribed to “The Skeptic AI Enthusiast,” you should have received a mail with my recent post, “Turing, the legendary AI prophet.” In case you didn’t spot the article or you didn’t get the mail, here is the link to it (free subscribers can read it up to a point, and full subscribers can read all of it):

Here is the end of the free preview of “The Skeptic AI Enthusiast.” Please become a paying subscriber for more exclusive content. Full subscribers can read the whole article I just mentioned, as well as the “Quotes” and “What is” sections, this one explaining what “Physical Intelligence” (discussed briefly in the news) is, and finally, a set of “friend links” for some curated articles on Medium so you can read them without having a Medium subscription.