Search Results for: #ai

I started looking into #AI almost 7 years ago. I started a really deep dive 5 years ago. I would have never imagined I would end up working with the former CTO of IBM Watson and his team, training AI models during and after working hours.

I am forced to pause my late-night experiments with #stablediffusion #AI to read this new theory of consciousness. I wonder what @danieldennett thinks about it. >

“Consciousness as a Memory System” https://t.co/ZttTmEzwhH https://t.co/3jIrsbHRZ1

Saturday 19 October 1889 (almost exactly 133 years ago), from the Oklahoma State Herald: press reports about an artist complaining about text2image #AI. I mean: complaining about instantaneous photography.

(found on Reddit. Original is here: https://t.co/2o0u8qUnte) https://t.co/n9DAdQu0Dy

Next challenge for #AI large language models: writing screenplays.

👀 https://t.co/Di4re5VqdV

Candid & best-intentioned Q to everyone:
What do you think is the difference between a person that looks for years at the code available on GitHub, learns from it, becomes a pro dev, publishes his/her own code, advises peers, writes books. etc. VS. an #AI that does the same?

I expect @StabilityAI to release more than text2image #AI models. If they trigger a collective participation in voice generation as they did for image generation, what we are seeing today in terms of creative output is nothing.

https://t.co/u6VIy7t9UG

In the last 2 weeks, I spent my free time working incessantly on #stablediffusion, exploring its potential and evaluating the myriad of forks that appeared online. Now we are so close to being able to fine-tune the #AI model on consumer computers thanks to Google’s Dreambooth.

I absolutely believe that #ai #deepfake tech will be used at scale to manipulate people. This is a must-read. https://t.co/5OmvMa8MXZ

I love stumbling on an unexpected, exceptional idea. I also love filing patents.

#AI

1/ The development of #AI image generators, especially for #stablediffusion, is going so fast that we now have the “prompt-to-prompt” technique: a user can request a modification of the generated image by simply writing what to change and how (rather than via in/outpainting).