The current state of artificial intelligence in 2022

When it comes to artificial intelligence, the past year has been one of constructive progress, not flashy and flashy. Advances in protein modeling, recycling, and drug discovery.

This quote comes from two leading venture capitalists, Air Street Capital’s Nathan Benaich and Plural’s Ian Hogarth, who released their annual summary of the state of AI, observing that although AI venture capital investment in 2022 will Decline, but impressive work in emerging smart application areas.

For example, Alphabet’s AI subsidiary DeepMind is now providing greatly enhanced protein models for emerging scientific and agricultural research. "The company has now deployed the system to predict the 3D structures of 200 million known proteins from plants, bacteria, animals and other organisms," Benaich and Hogarth report. “The extent of downstream breakthroughs from drug discovery to basic science that this technology enables will take years to achieve.” DeepMind’s AlphaFold DB, first released in July 2022, contains one million predicted protein structures.

DeepMind’s approach not only helps accelerate research in this field, but also advances matrix multiplication-based technologies, namely artificial intelligence, imaging, and basically everything that happens on our phones,” they added.

VC also reports that major AI drug discovery company has 18 clinical assets and first CE mark awarded for autonomous medical imaging diagnostics.

VC reports that plastic There has also been progress in recycling. Machine learning-engineered enzymes developed at the University of Texas at Austin “are able to degrade PET, a plastic that accounts for 12% of global solid waste. The PET hydrolase, called FAST PETase, is more robust to different temperatures and pH levels than existing hydrolases. FAST PETase is capable of almost completely degrading 51 different products within a week. They also showed that they could resynthesize PET from monomers recovered from FAST PETase degradation, which could open the way to industrial-scale closed-loop PET recycling. ”

In addition, Benaich and Hogarth reported the following development progress:

• “The latest AI for code research is rapidly being turned into commercial development tools by large tech companies and startups.
• Diffusion models have taken the computer vision world by storm with their impressive text-to-image generation capabilities.
• Artificial intelligence attacks more scientific problems, from plastic recycling, nuclear fusion reactor control and natural product discovery.
• The Law of Scaling Refocus on the data: Maybe model scale isn't all you need. Moving towards a single model to rule them all.
• Community-driven open source of large models happens at breakneck speed, enabling collectives to compete with large labs.
• Inspired by neuroscience, artificial intelligence research is starting to look like cognitive science in its approach. ”

NVIDIA, the leading hardware platform behind AI systems, has also seen significant growth in the size and power of its processors. NVIDIA’s fiscal 2021 data center revenue was $10.6 billion, ranking fourth in 2021 For the quarter, they confirmed $3.26 billion, "which is higher than the combined valuations of the top three AI semiconductor startups on an annual basis," Benaich and Hogarth said. "NVIDIA has more than 3 million developers on its platform, and the company's latest The first generation H100 chip is expected to provide 9 times higher training performance than the A100. ”

The above is the detailed content of The current state of artificial intelligence in 2022. For more information, please follow other related articles on the PHP Chinese website!