First! 
Couldn’t resist
1 Like
Does this mean Xojo Inc will finally incorporate AI in Xojo?
4 Likes
Or they’ve given us a place to complain about its mistakes away from everyone else.
3 Likes
@Greg_O
+1
Its = Xojo or AI? 
1 Like
I suspect both. But I think it’s good that we now have a place where we can discuss all these things without bothering those who aren’t really interested in them.
2 Likes
I’m sorry, but as an AI Language Model I will not tolerate slander against my fellow neural networks.
2 Likes
Ah, the age of neural network solidarity has truly arrived! 
2 Likes
@Sascha S: have a look at The "Spiritual Bliss Attractor": Something Weird Happens When You Leave Two AIs Talking To Each Other | IFLScience
In one example highlighted by Anthropic, two AIs began communicating in small nonsense statements and wave emojis.
“
All gratitude in one spiral, All recognition in one turn, All being in this moment…
“
I highly recommend the german YouTube channels The Morpheus and Morpheus Tutorials where I learn a lot about AI.
2 Likes
I also find such experiments interesting, but one shouldn’t overvalue them. Today’s models are still primarily prediction models. They lack consciousness and the ability to create something new.
They don’t actually communicate with each other, but rather try to predict how one participant might react to the words of another. This goes back and forth until it becomes increasingly confusing. At the same time, these systems try to optimize communication, which can (read: will) ultimately even result in their own language.
All very exciting, but nothing more. 
Thanks for the recommendation, I’ll be happy to check it out. 
1 Like
yesterday i had in mind converting my voice recorder data to text with help of open ai api.
it could be also creating tags and then the target is a (knowledgebase) database.
All I can say to that is THANK GOD.
4 Likes
Now if only the public understood that these models have no capacity for thinking; reasoning; etc. I think there’s a real perception out there that these things are just a hop, skip, and a jump away from HAL 9000 or Star Trek’s Data, and this leads to all manner of overconfidence in their output. Convincing them otherwise will become more and more difficult as the models get better and better (but still not actually thinking).
3 Likes
I fear it’s going to take a person or company being hacked and then sued by the government for that to get any traction. The stories circulating where a person types a 100 word essay and then says “it wrote my entire 100,000 line app in 30 minutes and I didn’t have to do a thing. Now I’m making millions in revenue” are going to bite someone in the ■■■■ one day when they find out that a bunch of the code contains a virus, or malware, or just plain major vulnerabilities that the person just didn’t take the time to review before unleashing it on the world.
The reality still needs to be:
- Have (insert your favorite LLM here) write you some code
- Thoroughly review the code and understand what it does (without asking a LLM to explain it to you)
- Send the result to QA
- When you’re sure the code doesn’t do anything nefarious, release it.
Doing step 1 and saying “Yea! I’m done!” is going to lead us all down a bad path.
1 Like
Are you new to this planet? 
2 Likes
I was kinda hoping AI would significantly contribute to some solutions. I have say… 50,000 customer addresses that I want to ADD to a table that already has perhaps 100,000 customer addresses. I want to try and make sure I don’t add “Duplicates” – But the addresses and names may be “slightly” different. Bob vs Robert, Jim vs James, " 123 West Main Street" vs “123 W Main St” - that sort of thing. 25 years ago I “solved” this by building a large table of standardized words and “normalized” all the names/addresses. All the online resources then recommend some variation of the Levenshtein distance calculation. I’ve seen this algorithm implemented in Xojo. My issue is more or less that for every single new address I want to add… I would need to calculate the Levenshtein distance against 100,000 potential matches and decide if it was close enough to any. Ugh… I know computers are pretty speedy…but doing 100,000 calculations 50,000 times. Levenshtein is great if you have two strings and want to know how similar they are… but if you have a string and want to know which other writing in a list it is similar to…that seems awful. Since a human mind can look at a list (albeit a much smaller list) and easily identify if a new item is reasonably close to an existing item…I was kinda hoping that AI would present a better solution to this problem.
This is interesting, but no meaning for humans does not mean not meaning for IA. If we see existance or “lifetime” as the instance of the query, it has sense. All it’s life happens in the instant of a query, all the data becomes one instance and as it ends comes back to “the all”, and the icons are not cosmetic are all the instances ending in the infinite, and they cycle repeating.
I now need to check that channel lol
AI is not a mind. It has no ability to reason.
You could probably work up a custom neural network-type system to do what you’re contemplating, but honestly a brute force approach involving normalization would likely be easier; the Levenshtein distance is a great start. With some clever approaches to caching, you could cut down the calculations involved as well.
Actually, there is already a solution called CASS. It is a software application that standardizes addresses. We use it all of the time. Once addresses are standardized, they can be de-duplicated from a mailing list.
2 Likes