Have we reached peak generative AI? • FRANCE 24 English

A slew of announcements from OpenAI and Google this week brought us to the brink of human-like AI assistants. Generative artificial intelligence is getting faster, and being applied to an increasing number of tasks. But without a major breakthrough to the models underpinning such projects, it has led some to emerge from the parapets and ask whether the technology will continue getting smarter given its exponential hunger for data. One of them is Dr. Michael Pound, Associate Professor of Computer Science at the University of Nottingham.
#AI #google #openai
Read more about this story in our article: f24.my/AL0a.y
🔔 Subscribe to France 24 now: f24.my/YTen
🔴 LIVE - Watch FRANCE 24 English 24/7 here: f24.my/YTliveEN
🌍 Read the latest International News and Top Stories: www.france24.com/en/
Like us on Facebook: f24.my/FBen
Follow us on X (Twitter): f24.my/Xen
Browse the news in pictures on Instagram: f24.my/IGen
Discover our TikTok videos: f24.my/TKen
Get the latest top stories on Telegram: f24.my/TGen

[@TobiPlayHD]

Dr. Michael Pound is a phenomenal educator that manages to captivate students and people interested in various topics around, e.g., computer science and machine learning. His videos on the Computerphile RUclips channel are incredible, as are the ones of the other presenters on there. Can’t recommend them enough.

[@clickrola]

Creating a perfect all knowing AI is like searching for the exact value of Pi you would never be able to get the exact value no matter how hard you try.

[@jptrainor]

ChatGPT told me that offshore oil rigs are resupplied using trucks. And it routinely totally screws up basic math.

[@99dynasty]

And don’t get mad, that’s actually your fault.

[@raydosson2025]

I feel like you definitely asked an older model

[@cagnazzo82]

You are likely using GPT 3.5

[@brandonreed09]

This is just the beginning. We are far from the peak.

[@cagnazzo82]

OpenAI releases a free model. And it is presumed that free model is the plateau?
Are we using logic here?

[@chrisso3082]

Not the plateau, but the claim is that the plateau may not be as high as many people seem to think. The claim they make is: The models improve in a different rate than the big companies claim they do (i mean i would exaggerate too if i can print money by doing it). And it totally makes sense that models won't magically recognize images or perform tasks where big datasets are not available. There are so many actions which could potentially be performed by AI but the data simply is not labelled or is too little or is too noisy...the list goes on. The next thing is benchmarking these systems is not straightforward and most of the time biased. I only know one website where actual experts rate the different models, and there the ranking is different than the one portraied by the various companies. So nobody knows for sure where the plateau is, but some make loads money by saying it will be AGI.