Mastodawn

rhempel Jul 17, 2025

@shanselman I saw this when you posted it on LinkedIn? a while back. The part where your voice catches in your throat as you describe your parents selling the van to buy that Commodore64 gets me right in the feels too ...

In the early days of the Internet we believed that giving kids computers would let them be as creative as we were when we got those first computers. Remember typing in code and saving it on cassettes, or the ill fated CueCat scanner :-)

We have lost a generation of creators to becoming media consumers, and set a stupid bar of success for "influencers" and monetization of everything.

Stop. Just stop. Connect with a true creator, someone that loves working with their hands, in wood, or leather, or on small engines.

Find those people one or two generations older than you and connect with creators that remember the time before the Internet and the enshittification that has followed.

Keep pushing back Scott, and remind us of the real promise that was made.

Alex White Jul 17, 2025

@shanselman

I understand the need to monetise videos, and I know that the decisions here are nothing to do with you, but it’s showing me an advert every 2-3 minutes, and that makes it REALLY hard to watch 😕

@PlaneSailingGames Eek! I will ask them, that's dumb. I didn't know it had ads

Paul Nicholas

Jul 17, 2025

@shanselman Watched it again and I get emotional every time 😭
It's both a wonderful glimpse of the past, and the kick up the butt to fight for the future that was promised 😅
Thank you for such a personal, honest and inspiring talk. Stay awesome, Scott - pls never change! 🤓❤️

Kevin Fox Jul 17, 2025

@shanselman I'd be far more likely to check it out if you could drop a hint as to the subject.

@kfury "**Tech Promised Everything. Did it deliver? | Scott Hanselman | TEDxPortland"**

mirabilos Jul 17, 2025

@shanselman we don’t want no stinking "AI" shit, why is this even in my timeline…

@mirabilos watch it...it's 90% anti

@mirabilos "**Tech Promised Everything. Did it deliver? | Scott Hanselman | TEDxPortland"**

@shanselman @mirabilos I’m surprised that your “good” use-case for AI is generating conversations with figures like Maya Angelou and Nelson Mandela. First, don’t you think there’s a consent issue there? Dead people can’t agree to have their work used for training, or to the creation of a fake representation of themselves. Second, the whole use case is abhorrent. Pretending that you can create an accurate simulacrum of a person based on their public writing/speech devalues their humanity.

@ripple @mirabilos that’s fair and valid feedback. I wonder if there’s validity in chatting with a “research librarian” with knowledge of a historical person’s public papers/books? I’m always trying to/open to reframing and rethinking my perspective

@shanselman what problem are you trying to solve? Do we need technology for this? How many research librarians/ teachers could you fund with the resources being spent on LLMs?

@ripple these are all great questions. Are there any valid used cases? I’m a bit of an AI vegan in that I don’t use it for images or video, and I try to use small models that are running locally and my own house. I don’t want it to eat the world. Is there any utility at all in your mind?

@shanselman I think the use of currently available LLMs is fundamentally unethical, so have not considered valid uses that much.
Emily M Bender has suggested that audio transcription might be a good application for LLMs, with appropriate safeguards.
Applications where the system “converses” with the users seem fraught because our brains are wired to assign intent to speech.

@shanselman in “The AI Con” Bender and Hanna suggest that we ask “what is being automated?” and “who will benefit?” This is difficult for technologists like myself, who have mainly been motivated by “what is cool?”

@ripple totally agree!

@shanselman how do small, local
models work? Do you start from scratch, or are they based on a pre-trained model that is then tweaked based on your training data?

@ripple depends. For my own nano model for my podcast it’s a distillation of a 3B llama trained on 1000 episodes of my podcast. For some of the Zulu and Xhosa stuff I do, I start with Vulavula which is ethically sourced on the continent. But yes, sometimes one uses an OpenAI model and turns it into a nano or pico model and the slippery slope begins

mirabilos Jul 18, 2025

@ripple @shanselman it’s also terribly TESCREAL to reduce humans to an autocompletion of their published works.

What makes us human is far more than that. Little conversations on the side, a smile, the treatment of others, …

felix (grayscale) 🐺Jul 20, 2025

@ripple @shanselman @mirabilos that part stood out to me too. "chat with an AI Jesus" doesn't seem to me like an improvement over "chat with an actor playing Jesus" except maybe the AI obscures the play-acting involved?

Arnaud Limbourg

Jul 18, 2025

@shanselman excellent talk, hits home on so many points.

Joeri Sebrechts Jul 19, 2025

@shanselman It wasn't tech that broke its promise, it was people. Technology is a catalyst, it accelerates whatever we apply it to. When we let people that want to go to a bad place decide where tech gets used, we get to the bad place faster.

vandys Jul 19, 2025

@shanselman When announcing a video, I suggest including what it's about? I see the hashtags, but maybe a sentence or two on the contents specifically?

Jason D. Moss 🇨🇦Jul 19, 2025

@shanselman

Wow. This is an excellent talk, Scott. Thank you.

Petr Stanislav Jul 20, 2025

@shanselman it was sooo great. Thank you!