Software coding and creation of algorithms in the connected ecosystem of the IT world has delivered the maximum for work, efficiency and for mankind in the last couple of decades. Though lots of progress was happening in AI yet the critical mass yet seemed to be far off. It all changed with ChapGPT hitting the public domain on 30th of Nov. 2022. The efforts of decades started coming to the fore in the form of versions, models and tools of AI, the impact of which we are facing on a day to day basis and the march goes on.
Coding and creation of algorithms are facing one of kind exponential changes which was not visualised for the decades to come. Though OpenAI stole the limelight yet lots of commendable work was happening at Google DeepMind for a long time and the results thereof have been impressing the tech and IT world, and lots of current AI growth and milestones have come out of that fold. Demis Hassabis, now a Nobel Laureate too, has been leading the charge directly for over one and half decades.
Now the clincher. Google DeepMind’s AI agent dreams up algorithms beyond human expertise. It is said that a new system that combines Gemini’s AI coding abilities with an evolutionary approach improves datacenter scheduling and chip design, and fine tunes large language models. The system in question is known as AlphaEvolve. The super-advanced AI can invent new algorithms. To prove what is being proclaimed, to its credit, AlphaEvolve has already made Google’s Data Centers more efficient and improved Tensor chips.
How does AlphaEvolve, essentially an AI coding agent, create the magic? It goes deeper than a standard Gemini chatbot. Due to the non-deterministic nature of the underlying technology, there is always a risk of hallucination. “AlphaEvolve uses an interesting approach to increase accuracy when handling complex algorithmic problem. This AI uses an automatic evaluation mechanism. When researchers interacts with AlphaEvolve, they input a problem along with possible solutions and avenues to explore. The model generates multiple solutions using the efficient Gemini Flash and the more detailed oriented Gemini Pro. At this stage each of the solutions is analyzed by the evaluator. It is this evolutionary framework that allows AlphaEvolve to focus on the best possible solution and improve upon.
ALPHAEVOLVE IS BY FAR THE BEST EXAMPLE OF MAN MACHINE INTERFACE’S PATH BREAKING RESULTS.
Sanjay Sahay
Have a nice evening.