OpenAI’s GPT-4 consistently beats The Turing Test

share to other networks share to twitter share to facebook
OpenAI GPT-4 Turing Test; SHODAN from System Shock Remake rogue AI looking menacingly

The Turing Test is one of the most infamous examples of AI intelligence. In a new development, it would appear that OpenAI’s GPT-4 AI may have completely bested the iconic anti-robot test.

Also known as The Imitation Game, the test is designed to see whether or not a computer could be indistinguishable from a human. If a human and robot were made to speak — or type — to an evaluator, would the robot be immediately caught out?

Advertisement

As it turns out, OpenAI’s newest language model AI may consistently best Turing’s test. But what does this mean for the future of artificial intelligence?

Reported by indie blog The Algorithmic Bridge, the new language model seems to have gone above and beyond the test. After years of having the test be the pinnacle of AI testing, the artificial intelligence software is convincing enough to be deemed a convincing human.

The news comes after OpenAI CEO Sam Altman tweeted a Star Wars meme that targeted the test. Altman’s meme proposes that even though the in-development AI is capable of beating the test, it’s still not as dangerous as the fictional powers of The Force.

At the time of writing, a number of AI softwares have beaten the test a few times. In fact, the current-gen language model GPT-3 has been used to beat the test before. However, it’s the consistency of beating the test that is the issue.

Judging from the responses from researchers working on the next-gen language model, GPT-4 is miles ahead of the current-gen model. This means that The Turing Test is now an outdated and obsolete test of the intelligence of, well, artificial intelligence.

Nowadays, there are more complex tests for judging the deceptive humanity of software. As The Algorithmic Journal notes, the Winograd Schema Challenge and The Embodied Turing Test are more modern, updated tests for 2020-era AI programs.

Advertisement

Read More: Anime fan uses $1,500 headset to date anime girl in IRL bar

Nevertheless, it’s impressive that we’re able to crush the test that the brilliant Alan Turing designed way back in 1950. Sure, it took almost 70 years to render the test completely obsolete, but it’s still the benchmark that many science fiction buffs will remember fondly.