TECH NEWS – AMD hardware powers the new OpenAI toolkit, and a minor glitch occurred during the “red” presentation, proving that you can’t always trust artificial intelligence.

AMD was also present at Computex in Taipei, and the company’s CEO, Dr. Lisa Su, tried to demonstrate how well its Radeon Instinct MI300X graphics card works with OpenAI’s tools. She presented Wanderlust, her travel assistant built on GPT-4. The only problem was that Wanderlust completely missed the location of Computex, although it did correctly find the city of Taipei… but it gave the previous location of the event, the Taiwan World Trade Center, as Changan Junior High School instead.

If AMD were trying to use the Wanderlust error to their advantage, they would say that they were just trying to show Taipei’s landmarks, but when the user started zooming in during the demonstration, the technology pointed to a fixed point. It shows that the demonstrations don’t always go as well as you might expect, but the twist in the story is that this time it wasn’t that the AI made a big mistake in a live broadcast, but that it was pre-prepared material and someone didn’t check whether Wanderlust was telling the truth or not.

And that’s why it’s embarrassing for AMD that Wanderlust screwed up, because in front of thousands of viewers it’s not a minor glitch, while the company presented it as the truth (when it was the opposite). This story shows that artificial intelligence should not be trusted 100%. By the way, AMD had other products at Computex, so the Ryzen 9000 CPUs and Ryzen AI chips were also on display, while the gaming theme was completely overshadowed in Nvidia’s presentation there…

Let’s hope AMD has learned from this little mistake, because this is no joke.

Source: PCGamer