Open AIis new GPT-4o Here it is — and it can laugh at bad jokes (and crack itself), sing along and help cheer up London cabbies with realistic emotions and regular human intervention.
OpenAI released today. 16 videos of GPT-4o (short for GPT-4 Omni) In action, the multimodal Foundation Large Language Model (LLM) interacts with the world in real-time male and female voices based on audio, visual and text input. shows
For example, after correctly identifying that the person it was talking to was preparing to make a big announcement—based on his professional attire and the presence of studio lights and microphones—the model was told that this was the subject.
A female voice answered, apparently friendly: “The ad is about me? Well, interesting me. You’ve got me on edge… Well, I don’t really have a set but you get the idea. Is.”
Open AI The new free model was revealed today at its highly anticipated Spring Updates event, with a remarkable 113,000 people tuning into the live stream. The model’s text and image input will roll out today in OpenAI’s API and ChatGPT, with voice and video available in the coming weeks.
Appreciating math skills, giving fashion advice
GPT-4o can identify users’ emotional state and environment, realistically simulate their different emotions and offer advice on many topics. Models on different devices can also interact with each other.
For example, in a video posted today by OpenAI, the model is described as interacting with another version of itself. To this, a female voice replies: “Okay, well, just when I thought things couldn’t get any more interesting – talking to another AI who can see the world. This is a plot twist in the AI universe.” Sounds like.
Asked to punch, direct and describe everything in their approach, the models took turns describing a man who is “sleek and stylish with his black leather jacket and light-colored shirt.” Tha was sitting in the “natural and artificial” room. Lighting that was “dramatic and modern” and a plant in the background added “a touch of green to the space”.
When a second person entered to give the first rabbit ears, GPT-4o was asked to sing a song based on what happened—and “playfully surprised the guests,” he groaned.
In other videos, the model laughs at dad jokes (“It’s absolutely hilarious”) translates from Spanish to English in real time and vice versa, sings a lullaby about “legitimate potatoes” (the answer to the first clue giving, “Now what’s that I call a mashup”), imitating a sarcastic voice like the droll MTV cartoon character “Daria”, Rock Paper-Scissors correctly identifies the winners and that That it is someone’s birthday is based on the presence of a piece of cake.
It also communicates with puppies—responding in a singsong tone the way people talk to dogs, “Well hello sweetie, what’s your name little fluffball?” (It was Bowser) — and guided a blind man through London, identifying via video input that the monarch was in residence based on the presence of the Royal Standard flag and the ducks described as “slowly wading.” Gliding across, moving fairly leisurely, not in a hurry.”
Additionally, GPT-4o can teach math. In one video, it shows a young man going through a problem based on an image of a triangle. The model asks the student to identify which sides of a triangle are oppositely adjacent and hypothesized, relative to angle alpha. When he guessed that alpha equals 7 to 25, the woman’s voice complimented: “You did a great job identifying the sides.”
GPT-4o can also give fashion advice. In another video, LLM helps a job candidate with scruffy hair wearing a slouchy T-shirt determine if he looks good enough for an interview.
A woman’s voice giggled and instructed him to run a hand through his hair. The model also remarked, “You definitely have the ‘I’ve been coding all night’ look down, which can actually work in your favor.”
Winning the Internet – or a Great Disappointment.
Not surprisingly, given the diversity of the AI community, the backlash, at least on social media, is all over the place.
Some are saying it “wins the Internet,” taking ChatGPT’s capabilities to whole new levels (and that it quickly competes with Google Translate). One user called the video teaching AI math “crazy,” adding, “The future is so bright.”
Nvidia’s senior research scientist Jim Fenn — among others — noted how the assistant was “alive and a little engaging,” recalling the 2013 sci-fi film “Her.”
Still others called it “the most underrated OpenAI event ever”.
Ultimately, AI consultant and investor Eli K. Miller commented, “Supertechies are frustrated that they don’t have a holographic laser beam coming out of their phone that reads their mind, and it seems that The wider business population wasn’t seeing it and weighing in.”
But that’s just one day of feedback—it’ll be interesting to see and hear the response when people get a chance to experiment with GPT-4o.
Credit : venturebeat.com
