A few useful weekend AI reads

Catching up on AI is a great way to spend part of your weekend

Intro to AI. This is the PowerPoint from my “Intro to AI & ChatGTPT” class I started teaching last week.

Sora and AIs that understand physics and the world. By now, many of you have seen the incredible life-like videos OpenAI is now able to generate with simple text prompts. If you haven’t check them out. This is a compilation from Echo Dude.

Many argue these videos prove that AIs can now understand (or are beginning to understand) the physical world, which constitutes a new level in AI development.

Many people missed the significance of the fact that both Sora and Meta's JEPA were released on the same day. Sora may demonstrate OpenAI's efforts to develop AIs in a way that understands the physical world. This is way bigger news than life-like video generation. It is something Yann LeCun has argued for a long time is essential for AI models to develop to get close to human-level intelligence.

This article reveals a bit of the difference in approaches and highlights this aggressive area of research.

As I’ve been saying, this is a big deal; AI isn’t about cheating on school papers.

AIs can do more than most people realize. One of the problems with ChatGPT3.5 being free is that people who use it think they understand what AI is capable of. But, 3.5 is only capable of text generation and it hallucinates a lot relative to more advanced AIs because it only has a fraction of its capabilities.

As Professor Marc Watkins lays out in the context of education, “Your Students Are Already Using AI You've Never Heard Of.

” This isn’t because your students are high-tech, it’s because they know to use more advanced AIs and different AIs for different purposes.

Rant. This isn’t a link to something to read, but it’s a brief rant: AI literacy for everyone is essential. In one of the educator Facebook groups I’m in, someone posted how they put their students’ papers in ChatGPT and asked ChatGPT if they wrote the paper. This isn’t the first person to do such a ridiculous thing, but faculty and staff need training so nonsense like this stops. Not only does the technology not work this way, but think about how absurd it is: How would ChatGPT know if another language model such as Claude, Pi, or many of the ones I run locally on my device wrote the paper? AI literacy in education is a borderline emergency.

AI is here to stay. Has anyone seen the NVIDIA stock returns?

NVIDIA is the company with the chips that power most of the AI tools.

Stock growth alone doesn’t prove value, and some say some of this is driven by hype, but remember we are just getting started. Even if some of the applications of AI are hype, imagine what happens when most people start using the tools for what they are already good at. This isn’t crypto, and pretending it is will leave you unprepared.

Embodied AI/Robots. As Jim Fan notes above, there is a lot of money to pour into continued AI development, including robotics and embodied intelligence. To learn more, check out this podcast.

<Let’s start with the theory: embodied intelligence posits that the body, or the physical form, plays a significant role in shaping an agent's mind and cognitive capacities. For example, human intelligence is not just the function of our brain, but a combination of our brain, our body, and the environment in which we exist. But when designing artificial intelligence (AI), a physical form and an environment are typically not part of the equation. It’s a disembodied cognition. Our guests, Li Fei-Fei and Surya Ganguli of the Stanford Institute for Human-Centered AI, set out to develop what they call an “evolutionary playground” to explore the development of embodied intelligence in AI and its connection with the environment and with learning using in silico experiments. They discuss with a16z general partner Vijay Pande and host Lauren Richardson how they created a suite of virtual environments in which agents evolve through a process that mimics aspects of Darwinian evolution. These agents, called the unimal, or universal animal, start off as a central node, and with each generation can add or subtract limbs and change various properties of their physical forms, like how flexible their joints are. Just like in real evolution, different forms arose based on the particularities of the environment, but what is really exciting is what Fei-Fei, Surya, and colleagues discovered about the intelligence encoded in some of these forms, such as an increased ability to learn a novel task. Which brings us to the applied section of our discussion. These results provide new insights for how we think about designing robots capable of performing unique tasks, and for understanding the possible limitations of disembodied AI models, like GTP-3.>

Our world is going to change.

This is from this video.