Alexander Wang: Building Scale AI, Transforming Work with Agents & Competing with China (AI column)

I watched something very interesting this weekend. speech Alexandra Wang, an American-Chinese entrepreneur born in 1997 One year in Los Alamos, New Mexico. In 24 He became the youngest self-made billionaire in the world. Wang is a mathematical prodigy who has participated in mathematics and programming Olympiads since childhood. On the ground. Scale AI He worked as a programmer at Quora at the age of 17 He studied at MIT, where he left to develop his startup. His company Scale AI is a data annotation platform that provides training data for machine learning models. (ala Mechanical Turk by AWS ortolokafrom Yandex). Alexander founded it with Lucy Go in 2016 A year in the Y Combinator Accelerator (Alexander was there then. 19). Scale AI was originally marketed as an “API for human labor,” but quickly focused on data for self-driving cars. (Alexander talks about this in detail in the podcast). Today, Scale AI is worth $29 Billions following Meta's recent investment of $US14 Alexander will lead the new laboratory of superintelligence Meta

If we talk about the main points of the speech, they are as follows: 1. The evolution of Scale AI: from data to agents Wang spoke about the company’s transformation from a simple data markup platform to an agent solution provider. The company has gone from focusing on self-driving cars to building hundreds of millions of dollars worth of AI applications for major corporations and governments. 2. The Future of Work: People as Agent Managers Alexander Wang presented a techno-optimistic vision for the future of work. The final state of the economy is “large-scale management of people and agents.” People will act as managers coordinating the work of AI agents, while maintaining control over vision and end results. It's funny that techno-pessimists sometimes say that agents will be people and they'll be managed by AGI:) 3. Competition with China in the field of AI Wang openly discussed the challenges of competition with Chinese AI labs. Chinese models are good partly because of industrial espionage, and China has advantages in data and energy. The US is lagging behind in energy production due to regulatory constraints, while China continues to build capacity. It was ironic to listen to a discussion of competition with China from Harry Ten, head of Y Combinator, and Alexander Wang, head of Scale AI. 4. "The Last Test of Humanity" Scale AI has created a benchmark called “Humanity’s Last Exam” – a set of highly complex scientific problems compiled by leading researchers. These tasks require hours of thought and have never appeared in textbooks. The best models are already gaining more 20Percentage of points, which shows rapid progress of AI in solving advanced research problems 5. Laws of Scaling and Specialized Models Wang stressed the importance of scaling laws, which became apparent with the release of GPT.3 into 2020 year. In the future, each company will have a specialized model as the main IP, trained on its own data and to solve specific business problems. 6. Agential workflows Scale AI actively uses reinforcement learning to automate internal processes, transforming human workflows into agency workflows. The company uses these solutions to analyze data, sales reports and other operational tasks.

In general, the interview is quite deep and interesting, and you can see how much Alexander is passionate about his work and is tinkering with the topic - it is not for nothing that Mark Zuckerberg invested in Alexander's company and called to lead the laboratory.

#AI #Engineering #Management #Leadership #ML #Software