Almost like the real thing

Artificial intelligence is all the rage, with machine and language models presenting all sorts of new tools for creators. Rev Fisk wrote last week about his adventures with ChatGPT and you can read about it here. But innovations in the use of A.I. are being released on a daily basis, it seems. 

Adobe will integrate generative AI models into its new Firefly suite, with users able to create images from text prompts. The company seeks to avoid copyright hassles of other models saying their machines have been trained on public-domain images, not pictures scraped from websites. Canva has also added generative tools to its app. 

Google announced its competitor to ChatGPTBard, but you’ll have to join the wait-list if you want to check it out.

Midjourney‘s new v5 is wowing users with its photo-realistic images generated from text prompts. But some complained that the near-perfect images didn’t give the dopamine hit that other generative models do – “the model’s precision takes away some of the thrill of repeatedly generating AI imagery to find a suitable result.” Oh, boo hoo. 

The rules around how machine learning can be used are still far from settled. Midjourney has allegedly banned Bellingcat founder Eliot Higgins from its platform. Higgins created images of an imagined apprehension of Donald Trump after the former president announced he was surely going to be arrested last Tuesday. Don’t worry if you can’t keep up, it’s all rather confusing.

The guys over at Not The Bee took Google’s new chat bot for a spin and found, as with others, that the algorithm is only as good as its training. On the political fractures and social issues of our day, there is a clear bias. We expect that if you put garbage in, you get garbage out, but as NTB point out, if this technology is going to have a big role in the future of mankind, this presents problems. 

An explainer on the difference between machine learning and deep learning.

Leave a Reply