Manuel Moreale reflecting on one hundred people who took the time to sign his guestbook (I did):

It’s the number of people who have taken a few minutes out of their busy lives to write something on my guestbook. One hundred doesn’t seem a lot in the grand scheme of things, especially on the web. If you have one hundred followers you might as well have zero. One hundred views on a YouTube video? That’s nothing. You need at least one hundred thousand to be part of the conversation. And yet, I find one hundred signatures in a guestbook to be a lot. Can you imagine having one hundred people in front of you, all saying something to you, one after the other? It would feel overwhelming.

I often think about this. Just like Manuel, I have a blog, a YouTube channel, a podcast (sort of). Globally, my follower counts are very low (even lower than Manuel’s) when looked through the Internet scale lens, but if those people were in front of me, in the same room, that would be quite something else. I find this thought to be comforting.

Great take by Gruber on this week sad but highly predictable events. The US is a sick country. Sidenote: on my last few trips, all Americans that I met told me after presenting themselves as Americans: “I know, we’re one of those stupid Americans…”. I reserve this thought mostly to Trump supporters which I never met personally, thanks god.

Warning: Unsettled thoughts: I think many tech pundits are overestimating Apple Intelligence capabilities and influence potential. If Apple fails to deliver, even slightly, it might trigger a crash like the dot com crash. Some tech pundits are fast at expecting Apple to be the gateway to generative AI legitimacy. In this logic, if Apple fails, AI will fail too. I might be over simplifying things here.

I’m wondering if the way someone elects to use generated content from generative AI models is way more potentially problematic and subject to debate than anything related to models training using content from the open web.

Also: generative AI content used to train generative AI models is also source of concerns to me. I call that process “knowledge desinfection” or “knowledge toxification” or even better “knowledge asphyxiation”. Or should I replace “knowledge” with “intelligence”?

One more thing: the more I think about generative AI training, the less I think it should be considered plagiarism. More on that one soon.

Thinking out loud about generative models training.

In a way, we’ve all been trained ourselves in our life by the books we read, the movies we saw, the music we listened. Some people have been trained on very specific knowledge bodies, in very specific fields. People use this accumulated “training” also forming “culture” to create new things and produce new content. Some people might be trained on a specific music style or dancing style. We’ve been trained by teachers. As “trained” creators, do we ask a permission when writing something new or writing music using our training data? Now because it happens at a large scale by large (and “nasty”?) corporations to create products, we cry foul?! Where is the line to be drawn here? I don’t know.