• FryAI
  • Posts
  • Altman says ChatGPT-4 “kinda sucks”

Altman says ChatGPT-4 “kinda sucks”

It’s thirsty Thursday, and that couldn’t come at a better time because our AI updates are salty! 🧂

(The mystery link can lead to ANYTHING AI related. Tools, memes, articles, videos, and more…)

Today’s Menu

Appetizer: Altman says ChatGPT-4 “kinda sucks” 😳

Entrée: Nvidia announces Gr00t for humanoid robots 🦾

Dessert: Google DeepMind showcases Tactic AI for soccer coaches ⚽️


🏈 Mysports AI: Use AI to help you with sports betting. → check it out

📝 Homework AI: Solve your homework with AI. → check it out

🗣️ InstaSpeak AI: Practice your English and public speaking in one minute increments. → check it out


If ChatGPT-4 sucks, then I must be a moron. 🙃

What happened? In an interview with Lex Fridman, Altman candidly remarked that the much-touted GPT-4 “kinda sucks.”

What else? Altman said that he anticipates the leap from GPT-4 to GPT-5 to be similar in proportion from the leap from GPT-3 to GPT-4. The GPT-4 model that Altman claims “kinda sucks” has propelled OpenAI to 100 million weekly users and an $80 billion valuation. But as great as GPT-4 is, its limitations are apparent, particularly in handling complex tasks. In this way, Altman's candid assessment serves as a reminder: innovation thrives on acknowledging shortcomings, driving a relentless pursuit for improvement.

“Look, I don’t want to downplay the accomplishment of GPT-4, but I don’t want to overstate it either … And I think this point that we are on an exponential curve, we’ll look back relatively soon at GPT-4 like we look back at GPT-3 now.”

-Sam Altman, CEO of OpenAI


Q: What is a robot’s favorite snack?

A: Micro-chips. 😆

What’s up? Nvidia unveiled GR00T (Generalist Robot 00 Technology), a general-purpose foundation model which is poised to transform how humanoid robots learn and interact with their environments.

How does it work? What sets GR00T apart is its ability to process multimodal instructions and learn from past interactions, translating them seamlessly into actions for the robot to execute. Powered by Nvidia GPU-accelerated simulation, GR00T empowers humanoid embodiments to learn complex tasks from just a few human demonstrations using imitation learning techniques. Additionally, by utilizing Nvidia’s Isaac Sim for reinforcement learning, GR00T can further refine its skills and generate robot movements directly from video data.

“Embodied AI will not only help address some of humanity’s biggest challenges, but also create innovations which are currently beyond our reach or imagination.”

-Geordie Rose, CEO of Sanctuary AI


Q: What position did the ghost play in soccer?

A: Ghoul keeper. 👻

What’s new? Through a multi-year collaboration with famous soccer club Liverpool FC, Google DeepMind has released what they call “TacticAI,” an AI-driven system that can advise coaches on corner kick strategies.

How does it work? TacticAI uses a proprietary model called Graph Imputer, designed in 2022 to analyze player movements and predict the probabilities of downstream effects in soccer matches. TacticAI builds off this model to allow coaches to sample different player setups and positioning and then directly evaluate the probable outcomes of such alternatives. According to Google DeepMind, the model addresses three core questions:

  1. For a given corner kick tactical setup, what will happen? e.g., who is most likely to receive the ball, and will there be a shot attempt?

  2. Once a setup has been played, can we understand what happened? e.g., have similar tactics worked well in the past?

  3. How can we adjust the tactics to make a particular outcome happen? e.g., how should the defending players be repositioned to decrease the probability of shot attempts?

What’s the significance? Strategy in sports is being driven more and more by analytics and probability. In fact, when this model was presented to soccer experts at Liverpool FC, they preferred the system’s strategies over human-formulated strategies 9/10 times. The potential of such a model is not only limited to soccer. Petar Veličković, a staff research scientist at GoogleDeepMind, says, “As long as there’s a team-based sport where you believe that modeling relationships between players will be useful and you have a source of data, it’s applicable.” As these models begin to find their way onto the sidelines, we can expect to see a seismic shift in the decision-making process of coaches across all sports.


What was your favorite headline from today's newsletter?

Login or Subscribe to participate in polls.


The Singularity Meter Rises 2.0%: NVIDIA releases their most powerful chip to date

What do ya think of this latest newsletter?

Login or Subscribe to participate in polls.