The most recent iteration of the technology behind OpenAI’s AI chatbotChatGPT has been made public. GPT-4o is the name of the feature that will be made available to all ChatGPT users, even those who are not subscribers.
It has been programmed to seem conversational and occasionally flirty in answer to commands, and it is speedier than previous models.
The updated version has the ability to read and analyze photos, translate between languages, and deduce emotions from facial expressions. Additionally, memory allows it to remember earlier cues.
Its conversational cadence is easier and it can be interrupted; there was no pause between asking a question and getting an answer.
About The Glitches
In a live demonstration, the voice version of GPT-4o offered practical advice on how to approach solving a straightforward problem written on paper, as opposed to just solving it. It translated between Italian and English, examined some computer code, and deciphered the feelings in a man’s smiling selfie.
It asked its prompters, in a friendly American female voice, how they were doing. It said, “Stop it, you’re making me blush!” in response to being complimented.
It was not flawless; at one point it began solving an equation that had not yet been demonstrated after mistaking the smiling man for a wooden surface.
This inadvertently showed how far chatbots still need to travel before the bugs and hallucinations that render them unreliable and possibly dangerous are fixed.
But it does indicate the path that OpenAI is taking, and I believe that GPT-4o is meant to be the next generation of AI digital assistants—a kind of Siri meets Hey, Google meets Turbo—that can communicate in ways other than text or voice and remembers what it has been taught in the past.
In addition to the excited cheers and applause from the off-camera audience, if there was an elephant in the room it was the environmental cost of this technology.
We know that artificial intelligence (AI) uses more computer resources than typical computing jobs, and that this requirement increases with the sophistication of the AI. The topic of sustainability was not brought up during the evening.
Although some chatbots, such as Elon Musk’s Grok and Mustafa Suleyman’s Pi from DeepMind, emphasize the “personality” of their products, GPT-4o’s ability to respond instantly and slickly to a combination of text, audio, and images seems to put OpenAI ahead of the competition.
Read Also: Chatting with Dad: How One Man Preserved His Father’s Voice in a Bot
Naturally, all we have to go on right now is what the company says: it was their demo, well chosen and run by them. As ChatGPT goes out, it will be interesting to observe how GPT-4o handles the millions of users it already has at scale.
GPT-4o is “magical,” according to OpenAI’s chief technology officer Mira Murati, but the company will “remove that mysticism” once the product is released.
It is not magic or sentient; rather, it is smart programming and machine learning. This is an intriguing and poignant choice of words, even though the technology is quickly advancing and growing more convincing as a companion.
The usage of Apple goods throughout the presentation was noteworthy, even though there have been rumors of a relationship between OpenAI and Apple that have not yet been confirmed.
The event’s timing, which came 24 hours before rival Google was scheduled to unveil its most recent AI advancements at Google IO, an annual conference, was another jab at the company.
