In a month where Nvidia became a $5 trillion company and Albania's Prime Minister said his AI minister is "pregnant" with 83 assistants, here is your roundup of AI news.

Browser dominance

As with seemingly every month, OpenAI has brought out a new feature, this time it's a direct competitor to Google's search engine: Atlas. In Layman's terms, Atlas is ChatGPT as a web browser: instead of just asking it questions, you can ask it for a URL and, unlike ChatGPT, the designated website will appear. It comes equipped with browser history, an AI assistant in a side bar, and it even has agentic features where it will act on your behalf (upon request).
It's a significant move by the American AI company and marks a competitive move in the tech space as they race for our constant attention. Even prior to the announcement, research showed that more people are turning to AI for their web searches.


They also released a feature that allows you to interact with other apps like booking.com or spotify. Say, for instance, you would like it to make a Halloween playlist in Spotify, it will simply access the app and do it for you. 

Humanoid household robot

For the price of $20,000, you may never have to do household chores ever again. 1X have have just put their robot NEO up for preorder, as humanoid robots could potentially become more common in our everyday life. The internet is full of videos of robots such as the one made by Boston Dynamics doing backflips, but now, having an actual robot assistant is a reality. 

The hefty price tag can also be divided into monthly installments of $500. 



Meanwhile, their direct competitor Figure also released a new version of their humanoid robot, named Figure 03. The internet was a foreign concept when it was first released and people were skeptical – but how long until robots become the norm?



AI that rewrites its own code

Self-learning AI is a concept that many leading researchers have warned against: an AI that is capable of changing its own code to self-improve. However, researchers at MIT have released a new paper entitled "Self-Adapting Language Models", where that exact phenomenon occurs.
They state: "Large language models (LLMs) are powerful but static; they lack mechanisms to adapt their weights in response to new tasks, knowledge, or examples. We introduce Self-Adapting LLMs (SEAL), a framework that enables LLMs to self-adapt by generating their own fine-tuning data and update directives."
Many processes within AI aren't fully understood, and adding uncontrolled self-learning to systems may raise a few concerns. Governments are racing toward superintelligence, and are doing so by throwing vast amounts of money into the industry. Is this another sign that they will stop at nothing to achieve their goal?

Warfare revolutionised

Anduril's new helmet, EagleEye, will undoubtedly change warfare. Through this new AI system, Anduril is creating, in essence, a coordinated beehive of soldiers and drones navigating the battle field in a symbiotic manner. The technology allows soldiers and drones alike to see behind walls and identify humans; once someone (or thing) has spotted an enemy, the rest of the battalion will also know where it is located. This will allow for incredibly precise reconnaissance and team movement. 

OpenAI's internal financial project

Bloomberg recently reported that OpenAI are currently working on a secret internal project entitled Mercury. It involves roughly 100 ex-bankers from high profiled establishments like JP Morgan & Chase, Morgan Stanley, and Goldman Sachs to train their AI models. They are supposedly being paid $150 an hour to create prompts that will potentially lead to streamlining the data entry process, also taking away a role that junior bankers normally perform. 

Sustainable AI?

A story that I missed last month but may have enormous implications... One of the biggest critiques of Artificial Intelligence is its energy consumption, with nuclear reactors being built solely to power data centres. Boffins at the Politecnico di Milano and an international team have published a research paper that has the potential to impact the way AI's computational power is produced. Through photonic chips, they use light interference on silicon microchips inside of them, which enables it to harness the energy to power them. The move will likely play a role in autonomous vehicles and wearable sensors.

You can read the previous months' AI reviews here.