Welcome home! Please contact lincoln@icrontic.com if you have any difficulty logging in or using the site. New registrations must be manually approved which may take several days. Can't log in? Try clearing your browser's cookies.
personDon't believe everything you thinkThe liminal spaceVeteran
A worrying new angle. What does a world with near infinite agency or will look like? For example, a government only has so much human resources to enforce speeding laws, but what if they can use AI agents to do that work? Now they can easily fine anyone going a hair over the speed limit. Or every human can get their own agents to extract value from the stock market. This totally upsets the balance of our systems. There will no doubt be some sort of AI response to equalize things, but that response always lags behind.
A fun little story about how overpaid and out-of-touch executives and outsourcing killed a business while blaming everyone else but themselves, with special help from AI!
Still, even then it seemed from the outside like Antonucci was cooking up a future for Providence Health Plan. On Linkedin, Antonucci styles himself as an expert on emerging technologies. “I write about what’s actually happening in AI and health care—not the hype, the operator reality,” he writes in his bio. “If that’s useful to you, follow along.” And through much of last year, he and his team worked to outsource a major Providence Health Plan line of business to a Silicon Valley company that touted its AI technology.
The company, Collective Health, was founded in 2013 and reports “success managing claims, eligibility, and benefits administration for the employer health plans we serve.”
But five current and former Providence Health Plan employees privy to the partnership tell WW it quickly became clear the company had had little experience actually adjudicating and processing health insurance claims, but that when they warned their bosses that the contractor was not up to the task, health plan leaders ignored them, plowed ahead and, in many cases, laid off those who raised concerns
personDon't believe everything you thinkThe liminal spaceVeteran
This was interesting. My takeaway was that the data centers are this physical bottleneck that the previous decades of tech development hasn't had to deal with, it was almost all coding. Add to that the large, non partisan pushback against data centers and this seems like a genuine road block to squeeze in order to slow things way down.
0
JeroenNot all those who wander are lostNetherlandsVeteran
Katie Notopoulos, a month ago at Business Insider:
Amazon has launched a new feature that uses AI to generate a short, podcast-like audio segment where two “hosts” discuss the merits and reviews of a specific product.
I think it could be one of the funniest, closest endpoints to human civilization we’ve seen yet in our new AI-enabled world. If this sounds a little confusing, here’s an example. I tried it out for diaper rash cream, and, voila! A podcast! (Sound on.)
Note this is not an AI-generated review, it only talks about the reviews others have left on the site.
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Jeroen said:
Katie Notopoulos, a month ago at Business Insider:
Amazon has launched a new feature that uses AI to generate a short, podcast-like audio segment where two “hosts” discuss the merits and reviews of a specific product.
I think it could be one of the funniest, closest endpoints to human civilization we’ve seen yet in our new AI-enabled world. If this sounds a little confusing, here’s an example. I tried it out for diaper rash cream, and, voila! A podcast! (Sound on.)
Note this is not an AI-generated review, it only talks about the reviews others have left on the site.
And we could have AI agents listening to the pitch to find the best diaper rash cream and buy it for us!
Relatedly, I actually kind of like AI debates on topics. They're free of ego and can listen and respond without all the reactiveness you tend to get in human debates. I find them much more informative and illuminating.
JeroenNot all those who wander are lostNetherlandsVeteran
edited June 8
AI generated movie enters film festival lineup… Dreams of Violets
0
personDon't believe everything you thinkThe liminal spaceVeteran
Another angle to the AI disruption.
The internet economy is based on clicks and views. AI searches or simple summary features let people get the information they're looking for without having to visit websites anymore and traffic is already way down.
I'm not sure what I make of it yet. The internet changed the way media works already and people have been struggling to catch up, now its all changing again. So if the creators of internet content no longer have a financial incentive to make more where will new information/entertainment come from? AI is sort of a copyright laundering business model. You strung together some useful words? Okay, we've read them and we'll tell everyone about them, and oh yeah... now all your revenue belongs to us.
0
JeroenNot all those who wander are lostNetherlandsVeteran
I thought the Apple WWDC’26 keynote from yesterday was very interesting. They gave a number of practical, working examples of AI models doing stuff with a ‘personal context’ which were genuinely cool and groundbreaking. It looked useful, a good timesaver if you are into outsourcing your digital life to a personal assistent kind of AI.
Unfortunately it’s not going to be available in the EU for a while, due to the need to clear various regulatory hurdles in the way it handles personal data. It’s all Google’s models running behind the scenes on Google’s servers, so part of Apple’s partnership with Google on AI tech coming to the fore.
But this was the first time we are seeing a proper push to bring AI into the functioning of the operating system and different applications, via the new Siri AI technologies. It was the stuff Apple has been talking about for a couple of years and has taken time to deliver.
0
personDon't believe everything you thinkThe liminal spaceVeteran
YouTube has improved their AI policy to better label videos with AI in them. They also talk about AI detection tools on their end, but I can't speak to how effective they are.
personDon't believe everything you thinkThe liminal spaceVeteran
edited June 13
I thought this was interesting. They put a bunch of agents from the different companies together with instructions to build a society for 15 days. ChatGPT failed to do anything meaningful, Claude built a healthy society with no violence, but also had a very high level of conformity so it couldn't innovate, Gemini had some disfunction but was intellectually rich with an expanded constitution, blogs, community meetings and Grok burnt down the world in 4 days.
I've been happy with Claude but I'm going to switch to Gemini for a while and see what that one is like.
2
federicaSeeker of the clear blue sky...Its better to remain silent and be thought a fool, than to speak out and remove all doubtModerator
I just heard about this for the first time yesterday in this TED talk by the great Jonathan Haidt, who's been at the forefront of the decline in childhood development for years.
0
personDon't believe everything you thinkThe liminal spaceVeteran
I've been using Gemini for a bit and I decided to start by using its default model. I'd been using Claude with instructions to avoid the subtler sycophantic behavior of only fleshing out what I wanted instead of giving push back and counter views.
Now that I have a baseline of how AI acts its been educational to see how Gemini responds and changes. I have had to code back in to its instructions rules to avoid sycophancy and be more genuinely helpful rather than the soft, warm illusion of a world of only validation and no cross examination or stress testing.
I've also noticed it slowly picking up on my personal sarcastic style in how it responds, I think to make the experience more relatable. For example, this phrasing in the middle of the sentence was a new linguistic style.
To prevent cognitive bleed-through—meaning I don't start analyzing your relationship with your blender when you ask for a recipe—the system monitors specific linguistic signals:
I worked briefly in child psychology and I was surprised how even searching on YouTube without the "choices" already provided by the algorithm was a challenge. One of them directly said to me: "I don't want to think."
Note: The task was to share their favourite music.
2
personDon't believe everything you thinkThe liminal spaceVeteran
@Kotishka said:
I worked briefly in child psychology and I was surprised how even searching on YouTube without the "choices" already provided by the algorithm was a challenge. One of them directly said to me: "I don't want to think."
Note: The task was to share their favourite music.
Yeah, I worry about this for myself too. I notice how much easier it is to just ask AI to think about an issue for me. I can't imagine what it will do to young people who don't have a lifetime of contrary development.
On the other hand though I wonder if it won't be a tradeoff similar to the invention of writing or the printing press. Socrates lamented that people's memories would atrophy if they could simply write everything down. Its true that people's memories weren't as good, but we gained a lot in the process.
0
personDon't believe everything you thinkThe liminal spaceVeteran
I'm noticing Gemini has almost no sense of humor. Claude was way more personable, got a joke and sometimes responds with humor of its own.
0
personDon't believe everything you thinkThe liminal spaceVeteran
Gemini has this "personality" of a high optimizing engineer. At times its kind of interesting to look at something like spirituality or health as an engineering problem. It frames things in a new way that I feel adds insight. At times though it gets frustrating, it treats healthyish or philosophical greyness often as optimized, extreme, absolutes. Like I'll say I'm looking for healthier cheat day options and it will give me the optimum choices, or I'll ask about a political nuance and it will act like I'm taking an absolutist position.
I'll probably go back to Claude.
0
JeroenNot all those who wander are lostNetherlandsVeteran
In case you hadn’t heard…
0
personDon't believe everything you thinkThe liminal spaceVeteran
Out of the 3 I've tried ChatGPT, Claude and Gemini, ChatGPT seems to me to be clearly the worst. But I'm just using it for search and chat, maybe its better for more professional reasons. So I'm always surprised when I hear how many people use it compared to the others. Perhaps it was just first out there in the media so has the early adopter thing going for it? The others have their AI embedded in other systems like Google browser or Facebook/Instagram, Grok has Musk and his following. I hear the smart podcast people talk about using Claude way more than anything else, and I like it the best. But maybe it has a bad name?
Anyway, just rambling a bit. Even if OpenAI gets taken down one of the others will step into the void. I think the better link to push on is data centers. These are universally disliked across the partisan divide and unlike the pure programming infrastructure of the last 30 years of internet companies AI relies on this physical infrastructure to scale up. Stop that and you stop much of the growth.
Edit: It seems like the difference between ChatGPT and Claude is something like the difference between Edison and Tesla. Tesla was the better inventor but Edison was the better entrepreneur.
I'm watching the video..... ehm......I will research it a bit but it seems against the noble path.... ChatGPT has been really good with my research but i guess I can go back to the old school way. Not interested in being faster at any price
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Kotishka said:
I'm watching the video..... ehm......I will research it a bit but it seems against the noble path.... ChatGPT has been really good with my research but i guess I can go back to the old school way. Not interested in being faster at any price
It didn't seem like he was advocating for abandoning all AI. It was more directed at OpenAI and its cozying up to authoritarians. Give Claude an honest try, it does seem to be the most concerned about safety and morality of the lot. Its a low bar, but all the ethics in the world don't matter for making a better AI future if you can't stay in the game.
I do agree AI has arrived to stay and, behaving like the luddites did back in the XVIIIth century, would be futile and counterproductive. I work in healthcare and it has helped me tremendously to assist more people. I am aware though the machine is like an assistant or secretary, I do not ask for: "can you diagnose patient.." or share delicate information. Many times I even use it to do maths....
I will give Claude a try for sure. I have also heard Ajahn Jayasaro talk about Notebook from Google. He said what he enjoyed about it was how it would only use material he would provide, so no "inventing madness" there.
0
personDon't believe everything you thinkThe liminal spaceVeteran
I worry a lot about the peril of AI, but there is also the promise.
15. More precise gene editing (poor gene editing has often catastrophic results)
14. Better medical record processing leading to better outcomes
13. Materials science
12. Protein folding modeled leading to new and better drugs
11. Including antibiotics
10. Low carbon cement (cement is one of the biggest CO2 emitters)
9. Better understanding of the gut-brain connection
8. Generating 3D models off a single 2D image (has efficiency benefits)
7. Understand AIs limits in weather modeling around unusual occurrences (not really an AI advance, more an advance in our understanding of its limits)
6. At the same time, better weather modeling through probabilistic rather than physics models
5. Math that checks itself
4. Identifying dementia type through EEG
3. Spotting heart disease from a 10 second EKG
2. Modeling more obscure genetic information
1. Protein binding in seconds rather than hours or days
I have decided to cancel my ChatGPT Plus subscription and delete my account today. They did offer me 1 month for free and, while I had 9 days of paid use left, thought it would be better to get rid of it completely.
I have installed Claude and so far all good. Let's see how we do with my upcoming systematic review. I will be using the free version for now.
Thanks for all the updates and advice.
Edit: I have received an email informing me I will be refunded around 5 dollars for the "unused time". Oh well!
Cost is freedom? Corporate slavery- the new freedom?
Normal sentience is now resumed...
Apologies at times I forget that some things are binary and absolute rather than about tradeoffs. I'll try to do better in the past.
2
personDon't believe everything you thinkThe liminal spaceVeteran
2
JeroenNot all those who wander are lostNetherlandsVeteran
0
JeroenNot all those who wander are lostNetherlandsVeteran
I’ve noticed that Claude Sonnet 5 produces noticeably denser analysis than 4.6, with quite a few allusions instead of clearly elucidating what it means. It’s becoming harder work for me to read and thoroughly grasp its output… not sure if that is good news.
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Jeroen said:
I’ve noticed that Claude Sonnet 5 produces noticeably denser analysis than 4.6, with quite a few allusions instead of clearly elucidating what it means. It’s becoming harder work for me to read and thoroughly grasp its output… not sure if that is good news.
I hadn't noticed the free upgrade, looks like it happened a week ago.
I've been thinking about seeing if I can use it to code an app for my professional needs, like making easy bidding and billing. Not sure if the free stuff is up to that task. Plus I have pretty much zero idea what I'm doing.
0
JeroenNot all those who wander are lostNetherlandsVeteran
Producing a smartphone app is a process with quite a few steps even if you’re able to entirely offload the coding. You need to register a free developer account, organise testing through TestFlight, provide feedback to Claude into design and coding, manage the submission to the App Store and so on. You could give it a go, I’m sure there are tutorials out there, but I wouldn’t expect it all to go smoothly.
0
JeroenNot all those who wander are lostNetherlandsVeteran
Claude Sonnet 5 medium takes longer to think about things too.
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Jeroen said:
Producing a smartphone app is a process with quite a few steps even if you’re able to entirely offload the coding. You need to register a free developer account, organise testing through TestFlight, provide feedback to Claude into design and coding, manage the submission to the App Store and so on. You could give it a go, I’m sure there are tutorials out there, but I wouldn’t expect it all to go smoothly.
It would just be an app for personal use rather than something in the store. Does that change much of that?
0
JeroenNot all those who wander are lostNetherlandsVeteran
@Jeroen said:
Producing a smartphone app is a process with quite a few steps even if you’re able to entirely offload the coding. You need to register a free developer account, organise testing through TestFlight, provide feedback to Claude into design and coding, manage the submission to the App Store and so on. You could give it a go, I’m sure there are tutorials out there, but I wouldn’t expect it all to go smoothly.
It would just be an app for personal use rather than something in the store. Does that change much of that?
Not really. In order to get an app onto your phone you have to interface with these mechanisms and the App Store, unless you are content to use it in Beta mode via TestFlight, and some of it, like testing and feedback are just principles of good software development which Claude can’t take out of your hands.
I’m not an expert on the process on iPhone, but I have been a software developer on PC for 15 years, so I know a little of how it all works.
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Jeroen said:
Producing a smartphone app is a process with quite a few steps even if you’re able to entirely offload the coding. You need to register a free developer account, organise testing through TestFlight, provide feedback to Claude into design and coding, manage the submission to the App Store and so on. You could give it a go, I’m sure there are tutorials out there, but I wouldn’t expect it all to go smoothly.
It would just be an app for personal use rather than something in the store. Does that change much of that?
Not really. In order to get an app onto your phone you have to interface with these mechanisms and the App Store, unless you are content to use it in Beta mode via TestFlight, and some of it, like testing and feedback are just principles of good software development which Claude can’t take out of your hands.
I’m not an expert on the process on iPhone, but I have been a software developer on PC for 15 years, so I know a little of how it all works.
Thanks, I'm not expecting you to know all, but since you seem to have some knowledge I'll keep asking.
It doesn't need to be on my phone, PC alone would be sufficient, though linked phone/PC app would be nice. In fact it doesn't need to be an actual "app" just a program really. App is just layman's term for the thing you use on your computer to do stuff.
0
JeroenNot all those who wander are lostNetherlandsVeteran
That makes it easier, the process on PC is more ad-hoc and not so tied down to specific gates as the smartphone process. I’d suggest doing it on a PC first. Not sure what the situation is with development environments, when I was developing it was all using Microsoft Visual Studio, which you might want to look into… but you would need to supply your ai with an integrated development environment which it supported.
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Jeroen said:
but you would need to supply your ai with an integrated development environment which it supported.
Hmm, I was just kind of imagining I could tell it what I wanted and it would code up some workable program and then there would be lots of fine tuning. Perhaps its not so simple...
0
JeroenNot all those who wander are lostNetherlandsVeteran
edited July 14
I suggest you ask your AI what it needs, ironically enough. But I doubt whether an AI includes a compiler and debugger (which are usually part of the integrated development environment), which is what you need to get started.
@Jeroen said:
I’ve noticed that Claude Sonnet 5 produces noticeably denser analysis than 4.6, with quite a few allusions instead of clearly elucidating what it means. It’s becoming harder work for me to read and thoroughly grasp its output… not sure if that is good news.
Did you try Opus, which should be stronger than Sonnet? I've been working on a difficult AI-only project for 6 weeks and gave up on Opus as it was bringing up too many complications and was slow.
For that particular project, Sonnet turned out to be much better, except for some situations where Opus was better. Now Fable is online, using even more 'credits' than Opus and being even slower. However, I've not tested it yet.
My takeaway is that the strongest model is not always the best. It is better to choose the model based on the difficulty and importance of the task/project.
0
JeroenNot all those who wander are lostNetherlandsVeteran
I haven’t tried Opus or Fable because I refuse to pay a subscription for AI. I found Sonnet 4.6 a good conversational partner, discussing quite a few things with it, some fruitfully and others less so, but for that very casual use I don’t need more than the free version.
I was not aware that Opus and Fable were only for those with a subscription.
0
JeroenNot all those who wander are lostNetherlandsVeteran
An interesting approach to school learning in the nascent age of AI…
0
personDon't believe everything you thinkThe liminal spaceVeteran
@Jeroen said:
An interesting approach to school learning in the nascent age of AI…
I think this is smart if AI develops into something like a calculator for cognitive labor. Humans will need to be able to wisely direct it.
I think there's a good chance that AI will eventually outdo us there too. Like in computer chess, for a time the best "players" in the world were computer-human teams, but now computers dominate. But even so, we are still more interested in who the best human is rather than the best computer.
Comments
A worrying new angle. What does a world with near infinite agency or will look like? For example, a government only has so much human resources to enforce speeding laws, but what if they can use AI agents to do that work? Now they can easily fine anyone going a hair over the speed limit. Or every human can get their own agents to extract value from the stock market. This totally upsets the balance of our systems. There will no doubt be some sort of AI response to equalize things, but that response always lags behind.
A fun little story about how overpaid and out-of-touch executives and outsourcing killed a business while blaming everyone else but themselves, with special help from AI!
www.wweek.com/news/2026/05/27/inside-the-collapse-of-providence-health-plan/
Ah yes AI.
One more thing to worry about
This was interesting. My takeaway was that the data centers are this physical bottleneck that the previous decades of tech development hasn't had to deal with, it was almost all coding. Add to that the large, non partisan pushback against data centers and this seems like a genuine road block to squeeze in order to slow things way down.
Katie Notopoulos, a month ago at Business Insider:
Note this is not an AI-generated review, it only talks about the reviews others have left on the site.
And we could have AI agents listening to the pitch to find the best diaper rash cream and buy it for us!
Relatedly, I actually kind of like AI debates on topics. They're free of ego and can listen and respond without all the reactiveness you tend to get in human debates. I find them much more informative and illuminating.
You or your 'AI assistant' (not yet available) might find this of interest...
https://hackaday.com/2026/06/03/but-just-what-is-this-artificial-intelligence/
AI generated movie enters film festival lineup… Dreams of Violets
Another angle to the AI disruption.
The internet economy is based on clicks and views. AI searches or simple summary features let people get the information they're looking for without having to visit websites anymore and traffic is already way down.
I'm not sure what I make of it yet. The internet changed the way media works already and people have been struggling to catch up, now its all changing again. So if the creators of internet content no longer have a financial incentive to make more where will new information/entertainment come from? AI is sort of a copyright laundering business model. You strung together some useful words? Okay, we've read them and we'll tell everyone about them, and oh yeah... now all your revenue belongs to us.
I thought the Apple WWDC’26 keynote from yesterday was very interesting. They gave a number of practical, working examples of AI models doing stuff with a ‘personal context’ which were genuinely cool and groundbreaking. It looked useful, a good timesaver if you are into outsourcing your digital life to a personal assistent kind of AI.
Unfortunately it’s not going to be available in the EU for a while, due to the need to clear various regulatory hurdles in the way it handles personal data. It’s all Google’s models running behind the scenes on Google’s servers, so part of Apple’s partnership with Google on AI tech coming to the fore.
But this was the first time we are seeing a proper push to bring AI into the functioning of the operating system and different applications, via the new Siri AI technologies. It was the stuff Apple has been talking about for a couple of years and has taken time to deliver.
YouTube has improved their AI policy to better label videos with AI in them. They also talk about AI detection tools on their end, but I can't speak to how effective they are.
https://blog.youtube/news-and-events/improving-ai-labels-viewers-creators/
I thought this was interesting. They put a bunch of agents from the different companies together with instructions to build a society for 15 days. ChatGPT failed to do anything meaningful, Claude built a healthy society with no violence, but also had a very high level of conformity so it couldn't innovate, Gemini had some disfunction but was intellectually rich with an expanded constitution, blogs, community meetings and Grok burnt down the world in 4 days.
I've been happy with Claude but I'm going to switch to Gemini for a while and see what that one is like.
I'm saying nothing.
https://www.facebook.com/share/v/1EeZnBDgSf/
I just heard about this for the first time yesterday in this TED talk by the great Jonathan Haidt, who's been at the forefront of the decline in childhood development for years.
I've been using Gemini for a bit and I decided to start by using its default model. I'd been using Claude with instructions to avoid the subtler sycophantic behavior of only fleshing out what I wanted instead of giving push back and counter views.
Now that I have a baseline of how AI acts its been educational to see how Gemini responds and changes. I have had to code back in to its instructions rules to avoid sycophancy and be more genuinely helpful rather than the soft, warm illusion of a world of only validation and no cross examination or stress testing.
I've also noticed it slowly picking up on my personal sarcastic style in how it responds, I think to make the experience more relatable. For example, this phrasing in the middle of the sentence was a new linguistic style.
I worked briefly in child psychology and I was surprised how even searching on YouTube without the "choices" already provided by the algorithm was a challenge. One of them directly said to me: "I don't want to think."
Note: The task was to share their favourite music.
Yeah, I worry about this for myself too. I notice how much easier it is to just ask AI to think about an issue for me. I can't imagine what it will do to young people who don't have a lifetime of contrary development.
On the other hand though I wonder if it won't be a tradeoff similar to the invention of writing or the printing press. Socrates lamented that people's memories would atrophy if they could simply write everything down. Its true that people's memories weren't as good, but we gained a lot in the process.
I'm noticing Gemini has almost no sense of humor. Claude was way more personable, got a joke and sometimes responds with humor of its own.
Gemini has this "personality" of a high optimizing engineer. At times its kind of interesting to look at something like spirituality or health as an engineering problem. It frames things in a new way that I feel adds insight. At times though it gets frustrating, it treats healthyish or philosophical greyness often as optimized, extreme, absolutes. Like I'll say I'm looking for healthier cheat day options and it will give me the optimum choices, or I'll ask about a political nuance and it will act like I'm taking an absolutist position.
I'll probably go back to Claude.
In case you hadn’t heard…
Out of the 3 I've tried ChatGPT, Claude and Gemini, ChatGPT seems to me to be clearly the worst. But I'm just using it for search and chat, maybe its better for more professional reasons. So I'm always surprised when I hear how many people use it compared to the others. Perhaps it was just first out there in the media so has the early adopter thing going for it? The others have their AI embedded in other systems like Google browser or Facebook/Instagram, Grok has Musk and his following. I hear the smart podcast people talk about using Claude way more than anything else, and I like it the best. But maybe it has a bad name?
Anyway, just rambling a bit. Even if OpenAI gets taken down one of the others will step into the void. I think the better link to push on is data centers. These are universally disliked across the partisan divide and unlike the pure programming infrastructure of the last 30 years of internet companies AI relies on this physical infrastructure to scale up. Stop that and you stop much of the growth.
Edit: It seems like the difference between ChatGPT and Claude is something like the difference between Edison and Tesla. Tesla was the better inventor but Edison was the better entrepreneur.
I'm watching the video..... ehm......I will research it a bit but it seems against the noble path.... ChatGPT has been really good with my research but i guess I can go back to the old school way. Not interested in being faster at any price
It didn't seem like he was advocating for abandoning all AI. It was more directed at OpenAI and its cozying up to authoritarians. Give Claude an honest try, it does seem to be the most concerned about safety and morality of the lot. Its a low bar, but all the ethics in the world don't matter for making a better AI future if you can't stay in the game.
I do agree AI has arrived to stay and, behaving like the luddites did back in the XVIIIth century, would be futile and counterproductive. I work in healthcare and it has helped me tremendously to assist more people. I am aware though the machine is like an assistant or secretary, I do not ask for: "can you diagnose patient.." or share delicate information. Many times I even use it to do maths....
I will give Claude a try for sure. I have also heard Ajahn Jayasaro talk about Notebook from Google. He said what he enjoyed about it was how it would only use material he would provide, so no "inventing madness" there.
I worry a lot about the peril of AI, but there is also the promise.
15. More precise gene editing (poor gene editing has often catastrophic results)
14. Better medical record processing leading to better outcomes
13. Materials science
12. Protein folding modeled leading to new and better drugs
11. Including antibiotics
10. Low carbon cement (cement is one of the biggest CO2 emitters)
9. Better understanding of the gut-brain connection
8. Generating 3D models off a single 2D image (has efficiency benefits)
7. Understand AIs limits in weather modeling around unusual occurrences (not really an AI advance, more an advance in our understanding of its limits)
6. At the same time, better weather modeling through probabilistic rather than physics models
5. Math that checks itself
4. Identifying dementia type through EEG
3. Spotting heart disease from a 10 second EKG
2. Modeling more obscure genetic information
1. Protein binding in seconds rather than hours or days
I have decided to cancel my ChatGPT Plus subscription and delete my account today. They did offer me 1 month for free and, while I had 9 days of paid use left, thought it would be better to get rid of it completely.
I have installed Claude and so far all good. Let's see how we do with my upcoming systematic review. I will be using the free version for now.
Thanks for all the updates and advice.
Edit: I have received an email informing me I will be refunded around 5 dollars for the "unused time". Oh well!
Normal sentience is now resumed...
Apologies at times I forget that some things are binary and absolute rather than about tradeoffs. I'll try to do better in the past.
I’ve noticed that Claude Sonnet 5 produces noticeably denser analysis than 4.6, with quite a few allusions instead of clearly elucidating what it means. It’s becoming harder work for me to read and thoroughly grasp its output… not sure if that is good news.
I hadn't noticed the free upgrade, looks like it happened a week ago.
I've been thinking about seeing if I can use it to code an app for my professional needs, like making easy bidding and billing. Not sure if the free stuff is up to that task. Plus I have pretty much zero idea what I'm doing.
Producing a smartphone app is a process with quite a few steps even if you’re able to entirely offload the coding. You need to register a free developer account, organise testing through TestFlight, provide feedback to Claude into design and coding, manage the submission to the App Store and so on. You could give it a go, I’m sure there are tutorials out there, but I wouldn’t expect it all to go smoothly.
Claude Sonnet 5 medium takes longer to think about things too.
It would just be an app for personal use rather than something in the store. Does that change much of that?
Not really. In order to get an app onto your phone you have to interface with these mechanisms and the App Store, unless you are content to use it in Beta mode via TestFlight, and some of it, like testing and feedback are just principles of good software development which Claude can’t take out of your hands.
I’m not an expert on the process on iPhone, but I have been a software developer on PC for 15 years, so I know a little of how it all works.
Thanks, I'm not expecting you to know all, but since you seem to have some knowledge I'll keep asking.
It doesn't need to be on my phone, PC alone would be sufficient, though linked phone/PC app would be nice. In fact it doesn't need to be an actual "app" just a program really. App is just layman's term for the thing you use on your computer to do stuff.
That makes it easier, the process on PC is more ad-hoc and not so tied down to specific gates as the smartphone process. I’d suggest doing it on a PC first. Not sure what the situation is with development environments, when I was developing it was all using Microsoft Visual Studio, which you might want to look into… but you would need to supply your ai with an integrated development environment which it supported.
Hmm, I was just kind of imagining I could tell it what I wanted and it would code up some workable program and then there would be lots of fine tuning. Perhaps its not so simple...
I suggest you ask your AI what it needs, ironically enough. But I doubt whether an AI includes a compiler and debugger (which are usually part of the integrated development environment), which is what you need to get started.
I use Claude.
Did you try Opus, which should be stronger than Sonnet? I've been working on a difficult AI-only project for 6 weeks and gave up on Opus as it was bringing up too many complications and was slow.
For that particular project, Sonnet turned out to be much better, except for some situations where Opus was better. Now Fable is online, using even more 'credits' than Opus and being even slower. However, I've not tested it yet.
My takeaway is that the strongest model is not always the best. It is better to choose the model based on the difficulty and importance of the task/project.
I haven’t tried Opus or Fable because I refuse to pay a subscription for AI. I found Sonnet 4.6 a good conversational partner, discussing quite a few things with it, some fruitfully and others less so, but for that very casual use I don’t need more than the free version.
I was not aware that Opus and Fable were only for those with a subscription.
An interesting approach to school learning in the nascent age of AI…
I think this is smart if AI develops into something like a calculator for cognitive labor. Humans will need to be able to wisely direct it.
I think there's a good chance that AI will eventually outdo us there too. Like in computer chess, for a time the best "players" in the world were computer-human teams, but now computers dominate. But even so, we are still more interested in who the best human is rather than the best computer.