The launch of Safe Superintelligence by Sutskever, DOOM simulation on GameNGen, saving 4500 working years thanks to Amazon Q — the top 3 AI news stories of the week
Our latest AI Digest covers the biggest breaking AI news of the week. Anywhere Club community leader, Viktar Shalenchanka, comments on key stories.
#1 — Safe Superintelligence: a company without a product
A new company called Safe Superintelligence (SSI) has emerged in the AI market. Its founder is Ilya Sutskever, formerly of OpenAI. The three-month-old startup currently has just 10 employees and a website written in plain HTML. It doesn’t have a product yet, but it does have the promise and reputation of Sutskever, it recently raised $1 billion from investors, and sources say it is valued at $5 billion. The company plans to spend its money on computing power and hiring specialists. SSI aims to create smart and safe artificial intelligence. Will they succeed? We don’t know yet. But we do know how much a reputation can cost.
#2 — GameNGen engine simulates DOOM
DOOM, the first-person shooter game released back in 1993, has been run on everything: calculators, smart home modules, and even electronic pregnancy tests (yes, really). But it had never been run on diffusion models, until now. Google Research developed a gaming engine called GameNGen. It is a generative diffusion model, based on Stable Diffusion, that predicts the next frames of the game based on the player’s actions. The result is a version of DOOM generated on the fly at 20 frames per second. It is difficult to distinguish from the real game and, although GameNGen has significant limitations, it being hailed as a development that could democratize game creation. We look forward to new games!
#3 — Amazon Q saves 4500 working years
Andrew Jassy, CEO of Amazon, reported that using the GenAI assistant Amazon Q to streamline foundational software updates has saved Amazon 4500 developer working years! He didn’t provide much detail, but he mentioned that updating applications to a new version of Java now takes just a few hours instead of 50 developer days. Jassy also noted the accuracy of the AI tool, indicating that nearly 80% of the code reviews it generated were implemented without changes and that the upgrades yield more than $250 million in efficiency gains over the course of a year. It’s hard to know whether Jassy is exaggerating these figures, but his representations do suggest a commitment by Amazon to leverage AI in software development.
.png?auto=webp)
