This Week In AI: Microsoft Levels Up Copilot, OpenAI Upsets Her, NVIDIA soars

It sometimes seems like artificial intelligence is on an inexorable march to take over pretty much everything. Companies and governments are already using AI for everything from stock photos to meeting notes. But we may need another leap in technology to go even further.

Meta’s AI head, Yann LeCun, told the Financial Times in an interview published this week that the large language model technology many current AIs are built on will not be able to think like humans. LLMs, he said, have “limited understanding of logic . . . do not understand the physical world, do not have persistent memory, cannot reason in any reasonable definition of the term and cannot plan . . . hierarchically.”

Perhaps sensing that many people feel as though they’ve seen AIs reason and use logic, LeCun said that in fact it’s just a sleight of hand. “It certainly appears to most people as reasoning — but mostly it’s exploiting accumulated knowledge from lots of training data,” LeCun said. Hence the importance of AI in knowledge-based tasks like coding and data analysis. 

Regardless, there’s still a lot that happened in the world of AI this week.

Microsoft upgrades Copilot

The software giant announced a new initiative called Copilot+ PCs, which are powered by a new version of Windows 11 built with the virtual assistant in mind. 

Copilot+ PCs will initially be powered by Qualcomm’s Snapdragon X Elite chips, which include neural processing units (NPUs) that can run local versions of Microsoft’s Copilot AI on the device, without an internet connection. Microsoft said it will also be adding Copilot into many aspects of the computer, whether it’s drawing images, writing emails or translating videos from another language in real time. 

The new Copilot+ PCs will include the company’s upgraded line of Surface tablets and laptops, as well as devices made by Asus, Dell, Lenovo and others. They’ll begin hitting store shelves June 18.

OpenAI’s newest AI voice upsets Scarlett Johansson

You know something’s getting serious when threats of lawsuits start. And that’s exactly what happened with Scarlett Johansson this week, when she released a statement saying she’d turned down a request by OpenAI to provide a voice for ChatGPT as a pseudo-reprisal of her role in the 2013 dystopian AI movie, “Her.”

Still, when OpenAI announced its new GPT-4o, its AI responded in conversations with a sultry voice eerily similar to the superstar actress. OpenAI pulled the voice after she hired a lawyer, but insisted it had not intended to copy her voice. 

Tech heavyweights promise AI safety

World governments and leading tech companies made a series of pledges at an AI safety summit in South Korea, promising to fund more research, testing and increased safety. 

Amazon, Google, Meta, Microsoft, OpenAI, and Samsung promised to steer AI away from bioweapons research, election meddling, or automated cyberattacks. They also promised to build a “kill switch” into their AIs in case any serious issues happen.

“We cannot sleepwalk into a dystopian future where the power of AI is controlled by a few people,” UN Secretary-General Antonio Guterres said in a statement. “How we act now will define our era.”

The U.S. State Department is investing heavily in AI

The Technology Modernization Fund, a program created by Congress to help fund technology upgrades and strengthen cybersecurity, has awarded $18.2 million to harness “safe, secure and responsible” AI “to empower its widely dispersed team members to work more efficiently and improve access to enhanced information resources.”

“This investment marks a pivotal moment to reshape how DOS conducts diplomacy and promotes American interests worldwide,” said Dr. Matthew Graviss, DOS Chief Data and AI Officer. “This project will enable our diplomats to focus on essential analysis and strategic engagement, boosting our efficiency and strengthening America’s global diplomatic advantage.”

NVIDIA’s AI chips are selling like wild

You may think of NVIDIA as merely the green-color-branded GPU maker, but its chips are at the center of the AI trend, and in a big way. The company announced more than $26 billion in revenue, driven by a 427% increase in its datacenter chip sales from the previous year. NVIDIA also said it expects sales to grow even more. 

NVIDIA’s shares topped $1,000 for the first time in response.