Keep pulling the thread on Risto Miikkulainen.
Researchers at Cognizant AI Labs, the University of Oxford, and NVIDIA have demonstrated that evolution strategies can successfully optimize neural networks with billions of parameters.
Sakana AI developed a system that autonomously initiated a problem, designed experiments, and wrote a research paper which was accepted by a major AI conference like ICML.
Cognizant, a company with 350,000 employees, is strategically shifting from a staff augmentation and consulting firm to an AI-focused company.
Risto Miikkulainen predicts that the next major trend in AI, following LLMs, will be agentic AI and multi-agent systems.
In the fall of 2021, Risto Miikkulainen's research group used their AI model to provide pandemic response scenarios and suggestions to the government of Iceland.
Risto Miikkulainen believes the future of AI lies in fostering creativity, a core strength of evolutionary optimization techniques.
The NEAT algorithm is considered robust and easy to apply to new tasks without requiring extensive parameter optimization.
The Evolutionary Computation Conference hosts a competition for "human competitive results" where AI systems routinely produce solutions that surprise and outperform human designers.
Risto Miikkulainen's research group has applied evolution strategies to fine-tune large language models such as GPT-4 and Llama after their initial pre-training.
In a recent Alpha Arena competition for stock trading, an unnamed "mystery model" outperformed all other participants.
Sentient, the company that preceded Cognizant's AI lab, developed a successful application of evolutionary computation for financial trading.
Google DeepMind's AlphaEvolve project is a notable example of evolutionary computation being successfully applied by researchers from outside the traditional evolutionary computation community.