Tesla’s Humanoid Robot, OpenAI’s Preparedness Framework, and Rumors About GPT 4.5: Top AI News of the Week
Our latest AI Digest covers the biggest breaking AI news of the week. Anywhere Club community leader, Aliaksei Kartynnik, comments on the key stories.
#1 — New Humanoid Robot from Tesla
— Tesla has unveiled the second version of its humanoid robot, Optimus Gen 2. Judging by the video, Optimus Gen 2 can perform various tasks: walking smoothly, squats, manipulation of fragile objects such as eggs, and even dancing (with moves that many people would envy)! All of this is possible thanks to improved sensors of Tesla’s own production, reworking of the robot’s balance mechanism, and the enhancement of its tactile perception.
It seems that we are moving into a world where humanoid robots will complement us and, in some cases, even replace us. Just a few years ago, Boston Dynamics was the undisputed leader in robotics. In 2022-2023, however, we saw the emergence of some bright new players, including Tesla. Tesla has clearly shown that impressive progress in robotics is possible in a short time: it took only a few months to develop Optimus Gen 2! We all remember, though, that in Tesla’s initial robot demonstrations, a human double was used instead of a real Optimus robot, so we eagerly await public demonstrations, not just another video featuring Optimus Gen 2.
#2 — AGI Preparedness Plan from OpenAI
— A few weeks ago, OpenAI announced the formation of a team called "Preparedness," whose task is to reduce the risks of using highly-capable AI systems in the future. OpenAI has already published a first stage of its plan for the safe development of future AI models — the Preparedness Framework. Pursuant to this framework, the team will assess the risks of new models in the following categories: cybersecurity, CBRN (chemical, biological, radiological, nuclear threats), persuasion, and model autonomy. After the assessment, detailed reports will be generated for the board of directors, which will make decisions about the company's developing models.
There are several interesting points in this news. First, the board of directors can now overturn decisions to release models, which was not previously the case. Perhaps this reflects the influence of the new board assembled after Sam Altman's return to the company. Second, this level of evaluation and reporting once again reminds us of the seriousness of OpenAI's intentions regarding the early creation of AGI — Artificial General Intelligence: AI capabilities that are comparable to human intelligence.
#3 — Rumors about GPT 4.5
— Over the weekend, social media exploded with rumors about GPT 4.5, suggesting that we will see an updated version of the model from OpenAI before the new year. Some ChatGPT users began posting screenshots that supposedly confirmed that a new model is already being used in ChatGPT. Sam Altman quickly debunked this rumor, and some of the screenshots turned out to be either fake or born out of the hallucination of the neural network.
Researchers are noting improvements in the quality and speed of ChatGPT, which may be the result of some as-yet-unannounced upgrades in the current model in use. The rest of us are eagerly awaiting holiday gifts from OpenAI!