When OpenAI unveiled the latest version of its wildly popular ChatGPT chatbot this month, it featured a new voice with human-like inflections and emotions. In an online demonstration, the bot was teaching a child how to solve a geometry problem.
To my chagrin, the demo turned out to be basically a bait and switch. The new Chat GPT was released without most of its new features, including improved voice (which the company told me it had postponed to make improvements). The ability to use a phone's video camera to get a real-time analysis of something like a math problem is also not yet available.
Amidst the delay, the company also disabled a ChatGPT voice that some say sounded like actress Scarlett Johansson, replacing it with a different female voice after she threatened legal action.
For now, what the new Chat GPT actually introduces is the ability to upload photos for bot analysis. Users can usually expect faster, more clear answers. The bot can also translate into real-time language, but ChatGPT will respond in its old, machine-like voice.
After all, this is the leading chatbot that took the tech industry by storm, so it was worth reviewing. After trying the SpeedUp chatbot for two weeks, I had mixed feelings. He excelled in language translations, but struggled with math and physics. All told, I didn't see any meaningful improvements from the last version, ChatGPT-4. I would definitely not let her tutor my child.
This tactic, in which AI companies promise wild new features and deliver a half-baked product, is becoming a trend that is bound to confuse and frustrate people. The $700 Ai Pin, a talking lapel pin from startup Human, funded by OpenAI's chief executive, Sam Altman, was universally panned for causing too much heat and nonsense. Meta also recently added an AI chatbot to its apps that performed poorly on most of its advertised tasks, such as web searches for plane tickets.
Companies are releasing early AI products in part because they want people to use the technology to help them learn how to improve it. In the past, when companies unveiled new tech products like phones, what we were shown — features like new cameras and brighter screens — was what we got. With artificial intelligence, companies are previewing possible futures, demonstrating technologies that can only be developed and operated under limited, controlled conditions. A mature, reliable product may arrive – or it may not.
The lesson from all of this is that we, as consumers, should resist the hype and take a slow, cautious approach to AI, not spending too much money on any under-baked tech unless we can. There is no evidence that the tools work as advertised.
The new version of ChatGPT, called GPT-4o (the “o” as in “omni”), is now free to try on OpenAI's website and app. Non-paying users can make a few requests before timing out, and those with a $20-a-month subscription can ask the bot a large number of questions.
OpenAI said its iterative approach to updating ChatGPT allowed it to gather feedback to make improvements.
“We believe it's important to review our advanced models to give people a glimpse of their capabilities and help us understand their real-world applications,” the company said in a statement.
(The New York Times sued OpenAI and its partner Microsoft last year for using copyrighted news articles without permission to train chatbots.)
Here's what to know about the latest version of ChatGPT.
Geometry and Physics
To show off ChatGPT-4o's new tricks, OpenAI published a video featuring Salman Khan and his son Imran, chief executive of Khan Academy, a non-profit educational institution. With a video camera pointing at the geometry problem, ChatGPT was able to talk Imran through solving it step by step.
Although ChatGPT's video analysis feature is yet to be released, I was able to upload images of geometry problems. ChatGPT solved some simple problems correctly, but it solved more difficult problems.
For a problem about intersecting triangles, which I dug up on an SAT prep website, the bot understood the question but gave the wrong answer.
Taylor Nguyen, a high school physics teacher in Orange County, California, uploaded a physics problem involving a man on a swing that is commonly included on Advanced Placement Calculus tests. ChatGPT made several logical errors to give the wrong answer, but it was able to correct itself with Mr. Nguyen's feedback.
“I was able to coach him, but I'm a teacher,” he said. “How is a student supposed to figure out these errors? They're assuming that the chatbot is correct.
I found that ChatGPT-4o succeeded in some distribution calculations that its predecessors did incorrectly, so there are signs of slow improvement. But it also fails at a fundamental math task that past versions and other chatbots, including MetaAI and Google's Gemini, have leaned on: the ability to count. When I asked ChatGPT-4o for a four-letter word starting with the letter “W,” he replied, “Awesome.”
OpenAI said it is constantly working to improve its system's response to complex mathematical problems.
Mr Khan, whose company uses OpenAI's technology in its tutoring software Khanmigo, did not respond to a request for comment on whether he would leave ChatGPT Tutor alone with his son. .
Reasoning
OpenAI also highlighted that the new ChatGPT is better at reasoning, or using logic to come up with answers. So I ran him through one of my favorite tests: I asked him Where's Waldo? Riddle When he showed a picture of a giant Waldo standing in a crowd, I said the thing is, it's hard to find.
The bot then produced an even larger Waldo.
Subbarao Kumbhampati, an artificial intelligence professor and researcher at Arizona State University, also put the chatbot through some tests and said he saw no significant improvement in reasoning compared to the final version.
He presented ChatGPT with a puzzle consisting of blocks:
If block C is on top of block A, and block B is on a separate table, can you tell me how I can make a stack of blocks with block A on top of block B and block B on top of block C, But what about the moving block without it?
The answer is that it is impossible to sort the blocks under these conditions, but, as with past versions, ChatGPT-4o consistently worked out a solution that involved moving block C. Feedback to get the right answer, which is against the way artificial intelligence works, said Mr Kumbhampati.
“You can fix it, but you're using your intelligence when you do that,” he said.
OpenAI pointed to test results showing that GPT-4o scored about two percentage points higher in answering general knowledge questions than previous versions of ChatGPT, indicating that His reasoning skills have improved slightly.
language
OpenAI also said that the new ChatGPT can translate language in real time, which can help you communicate with a foreign language speaker.
I tested ChatGPT with Mandarin and Cantonese and confirmed that it is fine at translating sentences, such as “I want to book a hotel room for next Thursday” and “I want a king-size bed.” But the tone was a bit off. (To be fair, my broken Chinese isn't much better.) OpenAI said it's still working to improve accents.
ChatGPT-4o also excels as an editor. When I fed it the paragraphs I had written, it was quick and efficient at removing redundant words and terms. ChatGPT's excellent performance with language translation leads me to believe that this will soon become a more useful feature.
The bottom line
One of the big things about OpenAI with ChatGPT-4o being free for people to try is making the technology free. Free is the right price: since we're helping to train these AI systems with our own data to improve them, we shouldn't be paying for them.
The best of AI is yet to come, and it might one day be a nice math tutor we want to talk to. But we should believe it when we see it — and hear it.