Glorified Stochastic Parrots with powerful tools

I enjoyed reading this post by Ploum where they describe their own personal AI policy – essentially refusing to interact with anything produced using these Stochastic parrots and creators who take pride in it. I’ve been seeing more and more posts like these, recently.

The direct consequence of my choice is that I will refuse to interact with humans who take pride in making themselves a meat interface between me and the algorithm they’ve chosen to replace their free will.

Martin Fowler thinks out loud and shares his feelings about LLMs, on a similar note. He mentions that he chooses not to interact with them, and seems to be particularly annoyed by their pretense to be human.

One of my most successful life-hacks is to avoid people I don’t like or don’t trust. I decline to interact with them socially, and make a deliberate effort to avoid working with them too, even if they are doing much that is beneficial. I feel that hanging out with pleasant, capable people, the people with integrity, has made my life a far better one. Hence my visceral dislike of interacting with an LLM that’s not just making a pretense of being human, but also posing as the kind of human I walk away from.

A colleague (M) who has also recently started to use these AI via chat interfaces seemed tired of talking to bots that sound like arrogant humans. Another colleague (S) and I talked about this a few times, and last week he shared this talk exploring different ways that humans and AI could collaborate which had some cool ideas.

I had the opportunity to learn about LLM harnesses a little bit last week, and it helped me see even more clearly how the “AI” that we are currently obsessed with is just glorified text prediction machines to which we, as a society, decided it’s a good idea to give access to any and every tool out there! And then we make a hue and cry about them being a risk and breaching systems and attacking whatever it is that they are “attacking”. The point Carl makes here about simply treating AI like we treat other software would be sufficient to put an end to a bunch of the alarm all around.

Speaking of trusting them with access to anything and everything, it also blows my mind how comfortable we all seem to be to give away our data to these LLMs to let them save us some time. It seems like we haven’t learnt anything from giving away our data carelessly to the previous edition of big-tech companies and their ad networks!

Anyhow…

This in-class rant by an IISc professor resonated with me. These are the kinds of questions I grapple with often these days.

Saa’s newsletter’s private web archive

I setup the web archive for Saa’s private newsletter. She had not been comfortable making the writing available publicly online, but had been wanting a way to share links of the writing with people (she trusted). The latest edition needed to have links to the old parts of a series, and so we setup a private web archive using Netlify’s functions and Identity (auth) functionality. Not sure if any of the new readers actually went back to the old editions, but a private web archive now exists.

Breakfast with Mala

Saa’s writer friend Mala visited us over for breakfast on her way back home from Morocco. She seemed to enjoy the Dosa and Chutney as much as we enjoyed her stories! Mala and I also ended up having a connect through UD (Delhi/Mumbai Ultimate).

Workouts

I was only able to get in two workouts, but my cleans have started to feel better. But, I’m still awaiting feedback on them from my coach, before I celebrate.

Earworms left behind by friends

Ringo, Ash and the kids left on Monday. I shared a picture from the airport in my last weekly. They left behind a few ear worms for us, though.

• Tere Paas Main from Main Vapas Aunga
• Hum Tere Pyaar Mein Saara Aalam by Lata
• Lakdi Ki Kathi from Masoom
• and the most earworm-y, Wheels on the Bus by CoComelon