This is a July Reply in response to Yordi, a blog that is well worth a follow!
I read your article on AI Critics and felt like it missed some of the important facets of criticism so wanted to give just a quick response.
I get that AI is helpful; for work I use different LLMs for coding support as example. It’s also fun; as a non-designer I had a lot of fun with Midjourney in its first year. I occasionally will still generate images for my kids or a friend that asks as well (although I no longer subscribe to Midjourney).
I also think AI has huge potential to address accessibility concerns (whisper models with translation for example; or diffusion models for things like automatically adding alt-text to images) and smaller highly focused work tasks.
I do however try to limit use and voice some genuine concerns. In fact many of the critiques I’ve read might use the six-fingered hand (or the suggestion of adding glue to pizza) as a foible to a joke to try and get people to understand some of those concerns by setting the stage: AI isn’t the salvation it’s sometimes made out to be.
Back to the concerns - a lot of those concerns comes down to the ethics of the situation:
- Perplexity AI has highlighted how many outright lie about user agents to skirts web server rules as they scrape, digest and use content. (Link to the Robb’s post that spawned articles from Wired, Cloudflare and many others).
- Lying about who they are just scratches the surface though; there’s also the matter of how the models are trained (specifically whether consent has been given to use the materials and whether licenses are honored in dispersing it). I don’t think anything has highlighted the importance of this as much to me as this lawsuit against AI music generators; the video with the AI songs is kind of wild.
- Environmental ethics is also a big piece. As of Dec 2023, MIT Review established that generating an image used as much energy as charging a cell phone (Stable Diffusion XL is the baseline in that claim).
- There’s also the human cost concern that some have: will generative AI making images of people 20 fingers and 3 legs replace designers on the job?
If the big AI companies were pushing these concerns forward and figuring out ways of honestly addressing them, things would be in a better spot, I think. It’s important to note: There are things that could be done. The MIT Technology Review makes that clear: it noted how they found small, finely tuned models to be much more energy efficient (as example); I think models like this are also going to be important for focused work tasks (like I mentioned above).
Long story short: I’m not going to stop following someone for using (or even promoting) AI. As I said above, I do too sometimes. I’ll still share the critiques though (even if they start with making fun of six fingered people) as those conversations are necessary to get to a place where generative AI truly is a net benefit to life.
Hope that all makes sense!
Have a thought? Or comment? Reply on Mastodon!