OpenAI says natively multimodal GPT-4o eats text, visuals, sound – and emits the same OpenAI on Monday showed off GPT-4o, its latest multimodal machine learning model, making it partially available to both free and paid customers through its ChatGPT service and its API. "The big news today is that we are launching our new flagship model and we are calling it GPT-4o," said Mira Murati, CTO of OpenAI, in a …
> It will be interesting to see whether OpenAI allows customers to use tone and simulated emotion to drive purchases or otherwise persuade people to do things..
Err, see icon.
The main, if not the -only- use-case for so-called Gen-AI is in deceiving or manipulating people in one way or another.
The only thing Open about OpenAI is its attitude to potential abuse, so it's unlikely they would cut off their main customer base: Scammers, grifters and bent politicians
If you are one of those who thinks it is writing code for you (it isn't, it is plagiarising other code and suggesting ways that your code could be just like all the other code, good and bad, that has gone before.. It has not even read the datasheet for the microcontroller you are using, and it does not 'understand' the problem you are trying to solve) - You won't like this new GPT anyway - it's multimodal. What use are cat videos, irrelevant images and pornography to a programmer?
Unless you are making documentation I suppose.. It could be along the lines of "Ha Ha, this code was vomited into existence by AI. I have no idea wtf it is doing. Here's some cat videos and grumble flicks to try to explain the software design"
"What use are cat videos, irrelevant images and pornography to a programmer?"
Are you sure you're doing this programming thing right? Cat videos are essential - even antisocial code warriors like to see a cute pussy now and then.
"It will be interesting to see whether OpenAI allows customers to use tone and simulated emotion to drive purchases or otherwise persuade people to do things."
Wasn't there some mention of nsfw output a few days ago? Couple this with that, and Mistress GPT is just a few versions away.
The developments sound impressive, though I haven't seen anything of the demo yet, and I could see a chatty assistant being useful, particularly, say to elderly folks who might be put off by less friendly products. But the problem of hallucinations seems to remain - while that's still present, you'd never be able to rely on it, largely negating it's usefulness to elderly or vulnerable users who might be tempted to trust it's every word. Still, it looks like progress has been made.
you'd never be able to rely on it, largely negating it's usefulness to elderly or vulnerable users who might be tempted to trust it's every word.
Given the proportion of people of all ages who, over the last two decades or so, have come to believe one conspiracy fantasy or another that they've read on the Internet, I think it's a little unfair to focus on the elderly* here.
*Pointing this out for a friend.
Not necessarily related to OpenAI, but I've noticed this too. People that are (superficially) intelligent are falling for major conspiracy theories. There are now countless people who believe that Covid vaccines are a population control tool, that Russia's invasion of Ukraine is perfectly reasonable, and that the world is being controlled by a shadowy business cartel. One of them captains long-haul passenger jets, although I didn't push the case as to whether he believed in chemtrails or not.
While I subscribe to the fact that if you believe in conspiracy theories you're a moron, it's concerning that people who have demonstrated flawed analysis and decision making processes are in a situation where they might need to use the same skills in a life-or-death situation.
"I think it's a little unfair to focus on the elderly* here."
Not so much focussed as suggested as one group with which I am familiar who might find a chatty LLM useful, subject to its reliability, but I take the point.
As for the conspiracy nutjobs, I hadn't really considered them to be a specific group in their own right. Rather, I'd put them in the vulnerable category, on the basis of their being hard of thinking.
The sheer number and variety of them is very disturbing though. It always amazes me that they believe "the government" or variations thereof, is capable of such subtle, tangled plots when all the evidence suggests that "the government" is largely incapable of getting through a day without becoming embroiled in one inconvenient scandal or another. Or is that what we're supposed to think...
"GPT-4o presents new challenges for us when it comes to safety because we're dealing with real time audio, real time vision," said Murati. "And our team has been hard at work figuring out how to build in mitigations against misuse."
Wouldn't it be a good idea to "build in mitigations against misuse" before releasing it to world+dog?
The Register 
Biting the hand that feeds IT
Copyright. All rights reserved © 1998–2025
