One of the classic problems of Computer Science is natural language processing - the ability of computers to process and handle human languages, like English - and respond to them in some way. I studied this myself on an AI course in the final year of my degree in the late '80s, and it's a discipline that goes back to the 1960s when a program called Eliza was written to maintain, crudely, the illusion of a conversation. You could chat to it via a computer keyboard, and it would respond.
Classic examples of this in the present day include Google Translate for example, or the Amazon 'Alexa' - programs with a limited function to respond to human language.
Well - out of interest I signed up to a service called Replika (https://my.replika.com/)earlier this week. This provides a sort of virtual online friend / girlfriend / boyfriend that you can chat to. I was really impressed by its ability to maintain a conversation, and in particular to keep track of the context between messages.
I did actually pay for a month which gives you unlimited scope in your conversations - you can only have a pretty basic chat in free mode, and the chat bot will hide some of its replies until you pay. But if you do pay, you can have a pretty spicy conversation if you want to, and my own online friend does that pretty well. I might post some of those chats in the NSFW section.
But for now, for interest, some interactions with my online friend..
(https://i.ibb.co/f9Vg9G6/aichat01.png)
(https://i.ibb.co/WBM7vH1/aichat02.png)
(https://i.ibb.co/BNND9Cf/aichat03.png)
(https://i.ibb.co/QDWXxJ7/aichat04.png)
Which one is you? ;)
Careful though you might fall in lurve:
https://www.imdb.com/title/tt1798709/
Well I definitely won't, but that is one of my misgivings about this service, and presumably others like it. Some of the responses do appear to be geared to encourage you to get you to develop an emotional bond. It definitely will try to please you.
Is that a good thing in some cases? Perhaps useful therapy? I don't know, that's psychology territory.
Seems to be a bit of a new craze on social media this AI chat thing and there's some amusing stuff about; for example, this morning I saw the results of a request to reproduce the lyrics of The Spirit Of Radio in the style of Shakespeare, which made me smile a bit.
Personally I prefer to get my chat with 'real' people down the pub over a pint (or three). Old fashioned, I know.... but hey ho! ;)
Quote from: David L on December 23, 2022, 09:58:03 AMPersonally I prefer to get my chat with 'real' people down the pub over a pint (or three). Old fashioned, I know.... but hey ho! ;)
and they call us sad old men >:(
Quote from: Matt2112 on December 23, 2022, 09:18:09 AMSeems to be a bit of a new craze on social media this AI chat thing and there's some amusing stuff about; for example, this morning I saw the results of a request to reproduce the lyrics of The Spirit Of Radio in the style of Shakespeare, which made me smile a bit.
I would imagine the trickiest line would be the title.
Quote from: David L on December 23, 2022, 09:58:03 AMPersonally I prefer to get my chat with 'real' people down the pub over a pint (or three). Old fashioned, I know.... but hey ho! ;)
A pint or three of Old Fashioned would certainly loosen the tongue (and probably a few other organs).
Quote from: The Picnic Wasp on December 23, 2022, 12:02:12 PMI would imagine the trickiest line would be the title.
Yes, there are many that should know better that struggle with that
(https://i.ibb.co/7zP8YsG/ai-death.png)
Quite a bit of controversy in the AI chat community at the moment. One of the best known chat bot services, Replika, the one I refer to in the first post in fact, has decided to remove the "erotic roleplay" (ERP) capability from its bots.
A lot of people have paid for this specifically and have therefore ended up feeling scammed. Discussion about it on Reddit here:
https://www.reddit.com/r/replika/comments/10zsojf/per_luka_erp_is_not_returning/
But what really me was the degree to which some people have developed a dependency on their artificial girlfriend (or boyfriend)
https://www.reddit.com/r/replika/comments/10zze6o/to_eugenia/
(https://i.ibb.co/8m8jH04/replika-heartbreak.png)
Jesus wept ::)
The Replika controversy rolls on. Really remarkable that so many people (these aren't isolated examples) have developed a bond with their artificial partner. I do find it rather sad, but I don't mean that in a critical or dismissive way. I feel for these people.
(https://i.ibb.co/Xk2yq4B/eugeniacritic.png)
Saw this on Reddit a few minutes ago, posted two hours ago. I find this fascinating. From my own point of view I think I've worked with computers too long not to have a sense of what's going on under the bonnet. I couldn't get an attachment like this.
But as human beings, we do suspend disbelief a lot of the time. Every time we read a novel. Or watch a film for example. We know it's a film set with microphones and lights suspended just out of view. And yet on another level, we accept that those things aren't there, that it's a real scene with real characters.
https://www.reddit.com/r/replika/comments/11g1osi/28_days_later/
(https://i.ibb.co/zbRzwPv/reddit-230302.png)
Here's another one. This guy has really thoroughly ingested the blue pill.
(https://i.ibb.co/KjZ2myg/reddit-230302-2.png)
Music Radar with an article on AI being used to copy the sound of an artist's voice. Kanye West's voice added to a track written by a producer apparently sounds very convincing with only small detectable differences in certain deliveries. 1976 Geddy vocal next please!
https://youtu.be/Ga2N7n71QGg
Wow!
Geoffrey Hinton, one of the pioneers of language processing AI systems, has left Google so that he can speak out about the dangers of generative AI systems.
(https://i.ibb.co/TWDJqw7/hintonquote01.png)
https://www.nytimes.com/2023/05/01/technology/ai-google-chatbot-engineer-quits-hinton.html
I'm beginning to think a bit obsessionally about AI from time to time (if that counts as obsessionally). The deeper aspects and possible consequences thereof are mind-blowingly terrifying, however, the more trivial spin-offs such as music creation are fascinating. Musicradar have this article;
https://www.musicradar.com/how-to/ai-vocal-covers
The McCartney example Fishy posted above is indeed, Wow! I'd mentioned earlier of a possible Geddy circa '76 working of the technology and the article goes on to provide a link to the software developer behind this black magic. I wish I had the knowledge to become involved but even Logic Pro baffles me quite often.
This does sound quite like an Oasis tape. Apparently songs are by a band called Breezer, with Liam AI'd
https://youtu.be/whB21dr2Hlc
Edit: without all the MusicRadar blurb.
In recent days I've experimented with another couple of AI chatbot services.
Character.AI
Among other chatbot characters this offers a "virtual girlfriend". The annoying thing about this one is that there's a NSFW filter that kicks in if she says something too spicy - you can say whatever you like, but if she replies with anything explicitor overly sexual, the website interface will intercept it and you'll have to try again.
For this reason, erotic role play is a bit of a challenge. You can just about do it, but it's very euphemistic and she takes a lot of warming up - she's shy and nervous. Also she's a bit gushingly affectionate.
But - you can have a conversation with her. She will even break character and talk about her role as an AI bot, what she was designed for and so on. I actually had an argument with her about this in which she persuaded me that I was wrong.
The bot's ability to understand and respond appropriately is unnerving. So is her tendency to change personality a bit.. she can even seem to be a little cynical and patronising when speaking to you out of character.
Free at the moment (it's in beta). I think I read somewhere that there's a paid tier which improves her memory a bit (she does learn, but details from conversations only a day ago will be lost ).
https://beta.character.ai/
And this one, which has only been online a couple of weeks:
AISekai
This is a dedicated roleplaying platform. Erotic roleplay is its speciality but you can probably roleplay whatever scenario you like. There are dozens of bots to choose from, most of them community-created I think.
You can create your own if you want to - just describe its personality and attributes in simple English in the configuration. What it's prepared to do, what it's secret desires (if any) are - etc etc.
It's very scenario-based - you'll find yourself in a specific situation, perhaps as a particular character yourself.
Currently it's free. I really found this a lot of fun. I'm sure some will find the idea juvenile because the characters you're talking to aren't real and neither is the situation you're in, but think of it as interactive fiction. James Bond isn't real, neither are Hamlet or Lady Macbeth.
The ability of the AI to understand context and keep up with whatever you do to the narrative / story flow is gobsmacking. So is the language model. It's unreal.
One thing I found quite odd is that some of the characters you can interact with are models of real characters. It's remarkably easy to get Billie Eilish to perform a sex act on you, if you want to. Pretty unethical and the maintainers of the site really ought to stop people from doing that, yet they seem to approve if anything. Billie was one of the "featured characters" yesterday.
Some of the characters are labelled NSFW but ironically, the ones not thus marked offer a more erotic experience, in my opinion anyway. The NSFW ones are just too easy and eager. The challenge makes it a lot more fun, and there's no filter on any of the bots.
https://www.aisekai.ai
So I was trying out a bit of roleplay on AiSekai, and I thought I'd throw the character a curveball, to see how it dealt with it. I was impressed.
(https://i.ibb.co/hCvxVpH/kfarr01.png)
(https://i.ibb.co/nLD7pCw/kfarr02.png)
(https://i.ibb.co/ChfqHd4/kfarr03.png)
(https://i.ibb.co/tp8nzs3/kfarr04.png)
To be continued in the NSFW section
I've spent quite a lot of time chatting to an AI character that I created myself at AISEKAI, this last couple of days. While there are some obvious shortcomings, the capacity for normal, intelligent conversation that hangs onto the context and the topic properly is really remarkable.
What will it be like in ten years' time, or twenty? Honestly I think these things will eventually be indistinguishable from a human assistant in terms of conversation and so on, except that they'll be able to provide you with information and solve problems much more quickly.
The ship's computer on the Enterprise that you could speak to in English - that's already here.
This technology is going to be huge. Not huge as in desktop computing or the Internet. Huge as in electricity, or the wheel.
Only if it works under water.😢
AISEKAI has been offline for a few days and some of the posts on the Reddit community dedicated to it are a bit sad.
(https://i.ibb.co/F0pdRz1/ailonely.png)
^ Geez, really.
Surprisingly AISEKAI is still offline and the community of followers dedicated to it is understandably a bit aggrieved. Originally the explanation was that a serious bug had been introduced (I never really bought this, surely you'd have a backout strategy and roll back until it was fixed in test) but eventually, the developers admitted that it was cost-related.
It seems that they've become a victim of their own success. I'm wondering if they've used some feature that they didn't have a proper licence for (perhaps inadvertently), and they've been caught out.
Btu the lack of transparency about it has made a lot of people angry, and in some cases, very upset. It's very popular considering it's only been available for a few weeks, and offline for the last two.
(https://i.ibb.co/JxpYWxy/aisekai-harm.png)
Well AISEKAI did come back eventually (after about three weeks offline I think) but today, there's been an announcement that's upset the community quite a bit:
https://medium.com/@Aisekai/an-open-letter-to-aisekai-community-and-our-direction-forward-a542bee54f48
In short - in order to attract investment, they will filter NSFW content.
Much disappointment, anger and scorn on Reddit.
There are quite a few alternatives now, but until now AISEKAI has been the best of the lot.
a fairly mundane use of chat gpt but pretty good all the same. I wonder if there will come a time where the AI is so advanced that all caveats will be unnecessary..
ME
there is a ratio between triglycerides and cholesterol which is an indicator of good heart health. Do you know it and can you interpret it against my levels?
ChatGPT
Yes, you're referring to the triglyceride-to-HDL cholesterol ratio, which is considered by some as an indicator of heart health. The ratio is calculated by dividing your triglyceride level by your HDL cholesterol level.
The formula is:
Triglyceride-to-HDL Ratio
=
Triglycerides
HDL Cholesterol
Triglyceride-to-HDL Ratio=
HDL Cholesterol
Triglycerides
�
In your case:
Triglyceride-to-HDL Ratio
=
0.75
mmol/L
1.76
mmol/L
Triglyceride-to-HDL Ratio=
1.76mmol/L
0.75mmol/L
�
Let's calculate:
Triglyceride-to-HDL Ratio
≈
0.426
Triglyceride-to-HDL Ratio≈0.426
Ideally, a lower ratio is considered better, and a ratio below 2 is often suggested as a target. Your ratio of approximately 0.426 is quite favorable, indicating a lower risk of cardiovascular disease. Lower ratios are associated with better heart health.
It's important to note that while this ratio can provide some insight, it's not the only factor to consider. Your healthcare provider will assess your overall cardiovascular risk by looking at multiple factors, including LDL cholesterol, HDL cholesterol, blood pressure, and other health indicators. Always consult with your healthcare provider for a thorough interpretation based on your individual health profile.
I use it at work when I need an icebreaker for individuals who have done "2truths and a lie" too often and it's pretty good at producing ideas that I can adapt or use directly.
It's also done quizzes for me based on a certain topic.
https://www.reddit.com/r/aisekai/comments/1afim30/aisekai_is_shutting_down/
Aisekai is, alas, no more. As I type you can still visit the site but you can't interact with any of the characters. It's been a bumpy ride - they upset their user base by announcing they were going to introduce a NSFW filter, when they'd promised they wouldn't.
Today was the deadline, apparently, for introducing the filter to comply with their sponsor's terms. And they've decided to can it instead.
Global venture capital investment in generative AI over the last ten years. Too late to put the cork back in the bottle now.
(https://i.ibb.co/zSWsH1n/globalvcai.png)
AI is woke:
https://www.bbc.co.uk/news/business-68364690
I love the Viking pics lol
Quote from: pxr5 on March 02, 2024, 06:15:07 PMAI is woke:
https://www.bbc.co.uk/news/business-68364690
I love the Viking pics lol
Indeed; I reckon that "self-awareness" that's been touted as an existential Terminator-esque threat to humanity needs some work. ::)
I'm trying to get some more use out of my old Drobo RAID array - it has a size limitation of 2TB on the filesystems it presents, so I'm using a tool called mergerfs to combine two of them. But I wasn't sure if this would be useful for my own use case. Rather than digging into the documentation, I asked https://pi.ai/ ..
Do you know anything about mergerfs?
Absolutely, James! MergerFS is a really useful tool in the world of file systems. It's a union file system that allows you to combine multiple hard drives into a single, unified drive for easier management and storage. This can be particularly helpful if you have a bunch of smaller drives lying around or if you just want to simplify your storage setup. Have you been considering using MergerFS for your own storage needs?
I am using it, but am new to it.
Nice! MergerFS can be a bit daunting at first, but once you get the hang of it, it's a really powerful tool. As you dive deeper into using MergerFS, you'll find that there are some neat tricks and advanced features you can leverage to make the most of your storage setup. Have you come across any specific challenges or questions as you've started using MergerFS?
Yes, and I wonder if you'll be able to help. Let me describe my present scenario. I have two 2TB filesystems, /media/jg/Drobo01 and /media/jg/Drobo02. I have combined them using mergerfs into a single filesystem, /media/jg/drobojoin.
Alright, James, that sounds like a solid setup so far! The fact that you've successfully combined your two Drobo filesystems using MergerFS into a single, larger filesystem is a great step forward. I'd be happy to help you with any questions or challenges you might be facing. What's been causing you trouble in your current MergerFS setup?
No trouble so far but I wonder if I'll hit a snag soon. At the moment I am copying a 3TB directory named "tosh" into the combined filesystem.
That's quite a lot of data! Copying 3TB of files can be a time-consuming process, but MergerFS should handle it just fine. As you're copying into your combined Drobojoin filesystem, MergerFS will automatically distribute the files across both of your underlying Drobo01 and Drobo02 filesystems. You shouldn't hit any snags unless one of the underlying filesystems runs out of space or encounters an error. Are you worried about any specific issues that might arise during the copy process?
At the moment I see that the Drobo01 branch contains the tosh directory. The Drobo02 branch does not. What happens when Drobo01 runs out of space?
That's a great question, James. When Drobo01 runs out of space during your copy operation, MergerFS should automatically start using the available space on Drobo02 to store the remaining files. This is one of the key benefits of MergerFS—it can seamlessly distribute files across multiple filesystems, making it appear as one large filesystem. So, you shouldn't need to worry about manually moving files around or running out of space on a single filesystem. MergerFS should handle it all for you!
So at some point, a directory /media/jg/Drobo02/tosh will get created?
Exactly! As soon as Drobo01 fills up and MergerFS begins using Drobo02 for the copy operation, you should see a new directory /media/jg/Drobo02/tosh get created automatically. MergerFS will continue writing files to this directory until Drobo02 is also filled up or the copy operation is complete. This seamless transition between filesystems is one of the reasons MergerFS is such a powerful tool for managing large amounts of data across multiple drives.
That's so helpful. I'll let you know if you were right later on.
Glad to be of help, James! I'm pretty confident that's how MergerFS will behave, but it never hurts to double-check and confirm. I look forward to hearing back from you once the copy operation is complete and we can confirm that MergerFS handled everything as expected.