Artificial Intelligence: Which should I choose?

Check a podcast by J Mosen on it. He uses one, do not recall what is the name but should be able to find it. From what I was listening sound it natural and was able to have a conversation. blindfully.

So, the app Jonathan Mosen was using on his living blindfully podcast was ChatGPT. He customised it, so it would answer to the name GP. But anyway, back to the original point. It depends what you want to do. I've used ChatGPT the most and Gemini too. However I've heard if you want to code, Claude is the best one according to the AI subreddits. I did get ChatGPT to write a python script to rename a load of files on my computer, and it did it! The script executed perfectly. I don't like copilot, it doesn't really do what I need it to. I'm paying for ChatGPT and Gemini, and I've paid for a month with Claude but it's no different to GPT and Gemini so I've cancelled my subscription for that. I haven't paid for Copilot though. Maybe the pro subscription is better? ChatGPT is pretty good at describing screenshots. If there is any text, it'll organise it into something coherent that you can actually read. Unlike straight forward OCR which tends to mix up the content on a page. GPT and Gemini are both pretty good at quizzing you on educational content you put through it. The only thing with both of them, is that if you feed it a really big document, it'll only ask you questions on the first 30 or 40 pages or so, even if you tell it to ask you questions on the whole thing. Writing and saving custom instructions for ChatGPT doesn't seem to help. Both Gemini and ChatGPT will throw in stuff that isn't in your original content you've either pasted or uploaded, and you sometimes have to remind it to stick to the content you've provided. I've tried out Claude with educational quizzes based on content I've been learning, and I didn't like the way it did it because it asked me questions in the exact order things appeared in the content, and I wanted it to mix things up a bit. I've never used Claude or Gemini to analyse images and screenshots and help me OCR stuff, maybe I should give those a shot too. ChatGPT is really good at rewriting and correcting content. I speak French and Spanish, neither of those are my native languages. However ChatGPT has corrected my writing, and I've asked a Spanish friend to have a look and he said it was good. I showed him a piece I had written with my mistakes, and then I showed him ChatGPT's corrections, and he said the corrections it gave me were good. About four or five months ago, I asked ChatGPT to edit an image. I had a screenshot of a page from a textbook, and I wanted it to remove all the text and leave me with the image. However I prompted it, it didn't work. It would remove some of the text but not all of it, it would rewrite the text in a cartoon-like font, so it just wouldn't do it. But last week, I tried it again. And it actually did it! I double-checked the result with AccessAI and a human, and only the image was left in the screenshot. It even gave me a .jpg file with the new edited screenshot I could download. The only thing I don't know, did it recreate the image or try and generate a new one, or did it actually remove everything I wanted from the screenshot, and leave just the image? Now to the not so great stuff. ChatGPT really isn't good at finding phone numbers and addresses on the internet even though it can search the internet. I asked it to find me some carpenters where I live, and it gave me a load of numbers of carpenters in other parts of the country! It has a new search feature which is meant to be similar to Google, but I asked it about an on-going news story I knew about that I had just been updated on, and it gave me inaccurate information. It told me it hadn't been resolved when it most definitely had over the preceding 24 hours! There is Venice AI which is uncensored. I've tried that, and it's pretty good. It can't describe images though. It can apparently generate images and code, but I've never tried that. There's mistral which apparently can describe images and correctly identify the characters in captchas, but I don't know how good that is. And I didn't even know about Rufus from Amazon, so I'll check it out. And I've tried Meta AI's chatbot but I haven't tried it enough to really comment. If you've got anymore questions let me know. I use AI for lots of things.

OK. Thought he also used another, I think he also show another.

Maybe Pi AI but I'm not sure. I know Double Tapp demoed that, but I'm not sure about living blindfully now. I wouldn't recommend Pi AI at all. I thought it was pretty bad. And I know Jonathan Mosen demoed Apple inteligence's writing features.

I use perplexity pro because you're paying the same price for the others individually, but they let you switch between about 5 AI Models.

I'm not sure if you can do the programming stuff, seeing as it's geared more towards search, but you can always try it for a month and cancel it if it doesn't work for you.

The unlabelled tabs use to be readable with screen recognition, but it looks like they broke it.

Home, discovery, and threads, or library is the 3rd one.

As for what AI models it supports.

let's ask it.

Perplexity Pro subscribers can access a variety of advanced AI models, including:

- **GPT-4 Omni** from OpenAI, known for its strong reasoning abilities and natural language processing, making it highly versatile and capable of performing human-like tasks.
- **Claude 3.5 Sonnet** and **Claude 3 Opus** from Anthropic, with Sonnet being an advanced model excelling in handling nuanced language tasks, and Opus offering a balance of speed and accuracy.
- **Sonar Large**, an in-house model built on Meta's LlaMa 3.1 70B, optimized for seamless integration with Perplexity’s search engine.

These models provide Pro users with a richer set of capabilities, allowing for more complex queries, creative writing, and deeper reasoning tasks. The choice of model can be tailored to specific needs, with Perplexity Pro users having the flexibility to choose the best model for their queries.

And what it didn't mention is now it also supports Grok-2.

There's a few things you can do with the pro that the free version doesn't give you.

Access to the other AI models being one, you only get the basic default with the free version, supposedly better responses to questions, oh and searching via voice.

I also use Perplexity Pro, and I'm planning to contact them with a couple accessibility related suggestions. It does work with programming as well.

I'm helping to test out Ally, which was formerly Envision Assistant. I really like it thus far. You can either talk to it or type to it. The web interface doesn't seem to contain much yet, but it's working great on my phone. It is free and very accessible with VO. It would be cool if they'd integrate it into their other app too on the phone, in addition to their glasses. I don't currently own those, but would really like to one of these years.

For me, I generally stick to ChatGPT. Claude is also pretty good, but on the free version, it's far too limited. The problem is, while there are many products on the market, most of them don't have a substantial free tear, similar to ChatGPT. I love using the model for creative writing. Not necessarily to write the content for me, but, sometimes, I do need a starting point. I'm not a professional author, and I'm not taking away anything from anyone else, so, yes, I'm free to do what I want, as long as I'm not hurting anybody else. I know a lot of people have very strong points about using artificial intelligence for creative writing. So, my thing is, I'll do what I want, and stay in my lane, and you can do what you want, and stay in your lane. Basically, nobody can come to my inbox, or comment section, with any nonsense. I don't have the time, temperament, or the patience for the foolishness. I don't let the model think for me. If I'm trying to figure out how to create a quantum 3-D printer for a supermarket for a fictional storyline, the models are great for coming up with the logistics of how the product could work in theory. Even if I'm not necessarily using the content, I love to be creative. For me, artificial intelligence, requires, from me, natural intelligence. It's not just about typing a prompt into the model and getting a storyline, or, parts of a storyline. If you don't know what you're trying to accomplish, then it makes no sense. This means that the model can essentially generate content that doesn't make sense, and if you put it out there, claiming that the work is yours, people are going to laugh you off the market. There is no idea that is new under the sun. We all draw inspiration from different places and people. The models package is the information in an accessible format to everyone. Sometimes, I ask about complicated topics, such as cryptography, genetics, stem cell theory, radiology, just for my own curiosity. Sometimes I generate pictures, videos, or to vent. Yeah, yeah, yeah. I know, I know. We're not supposed to replace human beings with robots, but, sometimes, I just want to talk, especially when I have complicated feelings or opinions about a situation, and I want someone to listen, without necessarily offering me their opinions. Sometimes, it's easier to talk with a robot, because there are no hurt feelings involved. I can't deal with too much emotions. Anyway, most of the models are trained on the same data, so, regardless of which ones you're using, you're more or less bound to get pretty similar answers, with minimal variations.

All of the models have their pros and cons. Unless you are doing a large amount of work and/or just want to be on the bleeding edge, the free tiers of most of these models can be quite adequate.

Here is a summary of some of the tools I use. There are web versions as well as iOS apps for each:

* Google's Gemini:
- Generally gives the quickest responses.
- Good for creative tasks like rewriting text, creating stories and poems, etc.
* ChatGPT and CoPilot - Both similar to each other
- Slower responses than Gemini
- Seems to provide more detailed, reliable, and accurate responses
* Perplexity
- This is my go to AI when I want a Wikipedia type answer that I can usually rely on
- Cites references for the pages from which it assempled its summary response. This is good if you want to verify the references and dig into more detail

And, as people already mentioned, there are many more. But these are the ones I use the most for different purposes.

--Pete

And of course, there's suno.ai for music creation. Using AI to create content is no different to an author having an editor or ghostwriter, or a singer having a producer write songs for them. The amount of music out there that isn't actually written by the artist themselves, or the amount of music where the producer instructs the band on what they should be doing, well I think we'd all be surprised by the statistics. and every book I've ever read is full of acknowledgements by the author, thanking all the people at the publishing house for editing etc.. So Winter roses if you were a professional author, you would have an editor help you with visual descriptions and getting things down on paper, this is no different to an AI giving you suggestions. You could go on a site like Freelancer or Upwork and ask for a human editor or ghostwriter to help you. Using AI is no different. People are trained on their own and other people's experiences. AI is trained on people's experiences. Getting back on topic, I tried copilot pro and wasn't impressed. It didn't do what I wanted at all. It seems to be more for news stories, and I don't need that. I've messed around with ChatGPT and creative writing too. It's interesting what it comes up with.