Hi, this is Allison Sheridan of the NosillaCast Apple Podcast, hosted at Podfeet.com, a technology geek podcast with an EVER so slight Apple bias. Today is Sunday, October 20, 2024, and this is show number 1015.
I’ve got a great show for you this week but it’s not exactly what I had planned. Last week you’ll remember that I had to use an AI-synthesized voice to read one of the articles for the show, and I promised you I’d talk about how I did it this week.
While I did write up an extensive article explaining the two services I worked with and the pros and cons of both, my voice isn’t yet strong enough this week to go through it. Don’t worry though, I do have one article and a follow-on conversation with Bart that will entertain and perhaps even educate you, and next week you’ll learn about how I created my fancy new laryngitis-replacing voice.
Articles
ScreenCastsONLINE Tutorial – Overcast
Let’s Talk Photography 133: Better than Your Eyes!
This week Bart was noodling what he should talk about on his next episode of Let’s Talk Photography and asked if there was anything I’d like to hear about. The answer I suggested turned into him inviting me onto the show to talk with him. The title of the episode is “Better than Your Eyes”, and it’s a perfect description of our discussion. We talked about how our phone cameras can now take better photos than we can even see. I learned this when viewing the recent Aurora photos posted online as people explained that the color wasn’t really there to their eyes. While it is possible to see the color, the Aurora needs to be quite bright and Bart explains why this is true.
Then we moved into talking about Steve and my new telescope, the Unistellar Equinox 2, which allows us to view and take photos of distant astronomical objects even from the light pollution capital of the world, Los Angeles.
As Bart and I always do, we had great fun sharing stories and even talking about the apps we use to track celestial objects in the sky.
Quick Use of ChatGPT to Write JavaScript for TextExpander Snippet
Support the Show
Our hero of the week is Keith Kubie. Maybe he heard about the cost of an AI voice for last week, or maybe he just realized he gets a lot of education and enjoyment out of the shows we produce here at the Podfeet Podcasts. Either way, he took his hard-earned dollars and took them to podfeet.com/patreon and signed up to support the show. I can’t thank Keith and all of the other Patreon members enough for their patronage. It makes me feel good and helps to pay for AI-generated voices when I get sick.
Plateaus Coming on LLMs – Discussion with Bart Busschots
Transcript of NC_2024_10_20
That’s going to wind this up for this week. Did you know you can email me at [email protected] any time you like? If you have a question or a suggestion just send it on over. Remember, everything good starts with podfeet.com. You can follow me on Mastodon at podfeet.com/mastodon. If you want to listen to the podcasts on YouTube, you can go to podfeet.com/youtube. If you want to join the conversation, you can join our Slack community at podfeet.com/slack where you can talk to me and all of the other lovely NosillaCastaways. You can support the show at podfeet.com/patreon like Keith did this week, or with a one-time donation at podfeet.com/paypal. And if you want to join in the fun of the live show, head on over to podfeet.com/live on Sunday nights at 5 PM Pacific Time and join the friendly and enthusiastic NosillaCastaways. Thanks for listening, and stay subscribed.
A fun episode. However, I disagree with Bart that LLMs are “mediocre at everything”, rather they present the illusion that they are mediocre at everything. And that is because of hallucinations. As I noted elsewhere, one needs to have an understanding of the domain to get useful information out of a LLM. In one prime example, from the MPU forum, a gentleman was making the argument of how much better ChatGPT was than Siri. He asked Siri multiple ways what was the exact time of the spring equinox in the US central time zone, and the best Siri could do was return the date. He then asked ChatGPT which returned an answer on the first try. The answer was wrong, but he was happily using that wrong answer.
LLMs may, as Bart claims, be better at Chinese than I am, but I have no way to verify that. For arguments sake, let’s say that an LLM is 95% accurate in a certain domain, but I have no knowledge in that domain. That means it will be wrong five out of 100 times, yet I won’t know which five. There is no way that could be used in a production capacity. That’s where the snake oil salesmen like Sam Altman and Elon Musk (although Musk is more of a fraudster) do a us great disservice. As you (Allison) showed, ChatGPT was helpful to you. And LLMs are useful to me as well. But the hype is overwhelming.
Gary Marcus reports and comments on AI developments. In his most recent email/post he ponders the question, “Is OpenAI more like WeWork or Theranos”, It is worth the read:
https://garymarcus.substack.com/p/is-openai-more-like-wework-or-theranos
There is a wonderful graph in there, I won’t spoil it for you.
With respect to changing the name from “Allison” to “Alister” and getting a different answer, researchers at Apple published a paper showing how adding a bit of extra, extraneous information, changes the answer, and for the worse. Here is a link to the paper:
https://arxiv.org/pdf/2410.05229
And knowing Allison’s distain for such things, here is the post by the aforementioned Mr. Marcus on the paper:
https://garymarcus.substack.com/p/llms-dont-do-formal-reasoning-and
Thanks again for a fun episode.
SteveM