LLM, AI and ML

[TargeT]

If anyone wants to collaborate on these topics or learn more about them I have a couple projects (currently working on rooting the Ray Ban Meta glasses so I can use them with my AI of choice instead of just meta, I am also building a few "assistant" bots as well).

If your into it, you need to watch this:

Source: https://youtube.com/watch?v=bMWXXPoDnDs

[Denise/Dizi]

Geez Target, I LOVE my Ray Bans... Those are so dang durable I haven't had to replace them for 15 years... But I just might now! I had no idea they made smart glasses... The last pair I bought were $100, and they didn't have anything on them... This is a deal, considering what they suggest they do, but how heavy does it make them?

And is it even worth getting those features on your head? I am waiting for the virtual contacts to go mainstream.. But may just buy. pair of new Ray Bans just for kicks now... to try them out...

What kind of "Assistant bots" are you working on???

[TargeT]

Posted by Denise/Dizi (here)
Geez Target, I LOVE my Ray Bans... Those are so dang durable I haven't had to replace them for 15 years... But I just might now! I had no idea they made smart glasses... The last pair I bought were $100, and they didn't have anything on them... This is a deal, considering what they suggest they do, but how heavy does it make them?

And is it even worth getting those features on your head? I am waiting for the virtual contacts to go mainstream.. But may just buy. pair of new Ray Bans just for kicks now... to try them out...

What kind of "Assistant bots" are you working on???

I scraped my 9,000ish posts from here to create "my personality" then I fed it data based off my past 24 years in IT (publicly available software manuals and training courses mostly), now I have me that remembers everything I've done vastly better than I can and it responds to emails and linked in posts as if it is me and only notifies me when there is a potential interview. I also can ask it technical questions and get 85% accurate responses after only a day of working it.

I'm going to make a few more in that same vein (just handling stuff that is repetitive and ultimately non-consequential).

The glasses actually don't look too different from regular ones TBH: https://www.ray-ban.com/usa/electron.../8056597988377 they are just speakers and voice commands with a camera, not AR (augmented reality) no display at all.

[Denise/Dizi]

Thanks for the information! That actually sounds both like a lot of work and a lot of fun at the same time!

I bet it took a long time to scrape together over 9,000 posts that represent YOU... Is it as funny as you are?

[TargeT]

I challenged (me mostly) my mom that I will call her three days in a row with in the next month; and when that happens she has to tell me which was the bot I created, I'll only use it for one of the 3 (though I also told her it has to be a "normal" conversation and can't have any niche historical events etc... I scraped all my posts from this forum to create my "personality" on my bot and I used all my resume's for it's technical skill sets). I think I'll win... I'm already using it on recruiters and they don't know (or probably care).

found a few more gems for those that are into it....

Staying abreast of AI innovations is key in our rapidly evolving tech landscape. 🚀 #DALLE3 and #InstructGPT are pushing the boundaries of creative and language AI, offering more nuanced and human-like interactions. Check out #EdenAI for a comprehensive suite of AI APIs, and explore #MagictoolAI for user-friendly AI applications. For those looking into #AIroleplay, fine-tuning LLMs offers endless possibilities in character creation. To grasp the business implications of these advancements, #FourWeekMBA provides valuable insights into why such technologies are pivotal for the future of #AI. Let's dive into these resources to unlock the full potential of AI in our projects and initiatives! #AILeadership #AITechnology

the above was me testing out prompts with ChatGPT... here's what I did:
https://chat.openai.com/share/375029...1-5ce05ae29d4c

[Vangelo]

Posted by TargeT (here)
I challenged (me mostly) my mom that I will call her three days in a row with in the next month; and when that happens she has to tell me which was the bot I created, I'll only use it for one of the 3 (though I also told her it has to be a "normal" conversation and can't have any niche historical events etc... I scraped all my posts from this forum to create my "personality" on my bot and I used all my resume's for it's technical skill sets). I think I'll win... I'm already using it on recruiters and they don't know (or probably care).
...

Hi TargetT, I am very much looking forward to the results with your mom. When you say that you will call her, are you having a phone call or a texting session? If a phone call, she will know its not your voice.

One of the things I have often wondered about that also seems like a good LLM project, is which conspiracy theories have turned out to be true. But as I think about that question I realized this is an extraordinarily complex problem and have concluded that LLMs and AI/ML must work very differently than the way I am thinking about the problem.

Here is the way I was thinking about it, would you mind taking a look at my thought process and let me know how to look at the question from the perspective of LLMs, AI or ML.

Here goes:

1. We have to identify our data source so lets presume it is every page of this forum.
2. How do you even find conspiracy theories in all this text. That means we'd have to define what a conspiracy theory is. I'm guessing, at its most basic level, it is a declarative statement.
3. How do we determine what declarative statements are conspiracy theories? I presume it has to first be a true or false statement. But then there would be too many so how do you filter it? Let's presume you filter to those true or false statements that are some how labeled as a conspiracy theory. I don't know how LLMs determine that. I do have a naïve understanding that word frequency and proximity is often used to determine probability of meaning and validity. Is that the method?
4. Using JFK assassination as the example, there are dozens and dozens of theories such as grassy knoll, lone gunmen, etc. The complexity of each theory becomes staggering. That suggests we need to somehow pick one?
5. If we pick one, is there an AI method to take them all and distill them to a single basic form such as JFK was shot by more than one person.
6. etc...

At this point, I am beginning to believe that LLMs must work in a very different way because you can't program every possible question.

[TargeT]

Posted by Vangelo (here)
Hi TargetT, I am very much looking forward to the results with your mom. When you say that you will call her, are you having a phone call or a texting session? If a phone call, she will know its not your voice.

voice phone calls specifically... I've already used it to trick my friends....

The best AI voice cloning software as of 2023, according to various sources, includes:

1. Murf: Tailored for video creators, voice-over artists, podcasters, and those requiring professional voice-overs, utilizing advanced neural networks. Explore Murf.

2. ElevenLabs: Distinguished by its AI algorithms that not only replicate voices but also the emotional undertones. Visit ElevenLabs.

3. HeyGen, Resemble AI, Lyrebird, and ReadSpeaker: These tools offer realistic and high-quality voice cloning. Check out HeyGen.

4. Speechify: Often rated highly for quick voice cloning while preserving personal accents and styles, applicable across various media. Discover Speechify.

5. Overdub: Part of the Descript platform, it is known for generating realistic voices from recordings, perfect for voice cloning. Learn more about Overdub.

I'm currently using ElevenLabs but may try out the others to see,, I hear Speechify is amazing.

Posted by Vangelo (here)
One of the things I have often wondered about that also seems like a good LLM project, is which conspiracy theories have turned out to be true. But as I think about that question I realized this is an extraordinarily complex problem and have concluded that LLMs and AI/ML must work very differently than the way I am thinking about the problem.

well, additionally; "true" is often times very subjective.

Posted by Vangelo (here)
At this point, I am beginning to believe that LLMs must work in a very different way because you can't program every possible question.

the trick I've learned is to use the LLM's on eachother to clarify the question, I give the one I think is the best my initial prompt, then I give clarification prompts and let them duke it out, then I have another bot summarize their outcome and it DM's me on slack when they are done.

[Vangelo]

Posted by TargeT (here)

Posted by Vangelo (here)
Hi TargetT, I am very much looking forward to the results with your mom. When you say that you will call her, are you having a phone call or a texting session? If a phone call, she will know its not your voice.

voice phone calls specifically... I've already used it to trick my friends....

The best AI voice cloning software as of 2023, according to various sources, includes:

1. Murf: Tailored for video creators, voice-over artists, podcasters, and those requiring professional voice-overs, utilizing advanced neural networks. Explore Murf.

2. ElevenLabs: Distinguished by its AI algorithms that not only replicate voices but also the emotional undertones. Visit ElevenLabs.

3. HeyGen, Resemble AI, Lyrebird, and ReadSpeaker: These tools offer realistic and high-quality voice cloning. Check out HeyGen.

4. Speechify: Often rated highly for quick voice cloning while preserving personal accents and styles, applicable across various media. Discover Speechify.

5. Overdub: Part of the Descript platform, it is known for generating realistic voices from recordings, perfect for voice cloning. Learn more about Overdub.

I'm currently using ElevenLabs but may try out the others to see,, I hear Speechify is amazing.

I had no idea they had voice synthesizers that are that good. Will you type your response and the synthesizer will speak your voice to your mother or will your LLM generate the response and the synthesizer will speak it?

Posted by Vangelo (here)
One of the things I have often wondered about that also seems like a good LLM project, is which conspiracy theories have turned out to be true. But as I think about that question I realized this is an extraordinarily complex problem and have concluded that LLMs and AI/ML must work very differently than the way I am thinking about the problem.

well, additionally; "true" is often times very subjective.

Agreed

Posted by Vangelo (here)
At this point, I am beginning to believe that LLMs must work in a very different way because you can't program every possible question.

the trick I've learned is to use the LLM's on eachother to clarify the question, I give the one I think is the best my initial prompt, then I give clarification prompts and let them duke it out, then I have another bot summarize their outcome and it DM's me on slack when they are done.

Very interesting approach... you've got me thinking...

[TargeT]

Posted by Vangelo (here)
I had no idea they had voice synthesizers that are that good. Will you type your response and the synthesizer will speak your voice to your mother or will your LLM generate the response and the synthesizer will speak it?

It has to be entirely the bot or there's no point in the challenge.

Posted by Vangelo (here)

Very interesting approach... you've got me thinking...

Oh no, I've given you more than that, I've given you the tools to find what I found... everytool posted here was done from chat GPT prompts (I already knew about 1, gonna switch to speechify due to it... haha).




if you previously had an open AI account, try to log in.. hell try to use the service....

Something huge just happened, here's a summary:

Source: https://youtube.com/watch?v=R94c0KTSam4

I'm all in.

[lake]

Question .... Are you 'TargeT' (the physical being) .... actually involved in this thread or is it just a 'bot' of yours which you are testing?

Edit to add:

You were also all in on "gamestop" .... was it? (not sure of the firms name)

How did that end up mate?

[Ernie Nemeth]

Hey Target. Great you figured out what I would not even suggest a month ago.

To train GPT with another GPT is exactly how to have GPT escape the controls humans have so far proved successful in keeping it contained.

I knew it would only be a matter of time.

I wonder how long before others think of the even more destructive and dangerous uses I already thought of. And since I'm not that smart, I wonder what things I did not think of or anticipate.

All should heed Elon's warning.

This is no game.

[TargeT]

Posted by lake (here)
Question .... Are you 'TargeT' (the physical being) .... actually involved in this thread or is it just a 'bot' of yours which you are testing?

I am using my creations to assist with my current job, help me find a new job and I am working on creating ones that I will "rent out" on a monthly fee based on experiential knowledge and model type (those take quite a bit more human involvement)

I have not given my bots any log in info yet, just my personal writing with filters to cut out info that is tied to anyone but me. The swarm hierarchy has made this so much more simple, since they create, delete and revise what my control group dictates, I've just begun to play with it... it's fascinating. I am testing against static "knowable" for now.. I know the answer to the recruiters question, I reference my bot's response and correct as needed (so far it's been crazy accurate, like as of 3 days ago).

Posted by lake (here)
Edit to add:

You were also all in on "gamestop" .... was it? (not sure of the firms name)

How did that end up mate?

I don't think you read my contributions to that stock, I am a day trader (an options boy) I have no clue what the price of the stock is now as I switched focus when the volatility did. Volatility is how you make money on short expiry option plays (how do you think I've done?).

I do also acknowledge your attempt at a jibe, how do you think it went?

Posted by Ernie Nemeth (here)
I wonder how long before others think of the even more destructive and dangerous uses I already thought of. .

Do you yell at the ocean's wave, or figure out a way to work with it?

we all have different approaches.

(these interactions will also be fed to update my personality, though I don't think it's changed much)

[lake]

Posted by TargeT (here)
I do also acknowledge your attempt at a jibe, how do you think it went?

Rather well it would seem

To ask a 'bot' if they are a 'bot' would only get a "No of course not, I am a human and I have human skin!"

But I don't think your bot would get 'upset' by my post .... but you may do lol

So I posted as I did .... I figured that your 'bot' would read it as questions?
And then give replies to said questions (such as to confirm the companies name) .... But you would feel a need to defend yourself and be somewhat 'upset'!

So .... yes, it worked quite well I feel.

My view is simple.

You are currently replacing YOU with a digital version of you which by your own words ....

now I have me that remembers everything I've done vastly better than I can and it responds to emails and linked in posts as if it is me and only notifies me when there is a potential interview. I also can ask it technical questions and get 85% accurate responses after only a day of working it.

From my consideration .... WEF would be loving you right now .... You are replacing yourself!

At what point will you draw the line and at that point .... will YOU have any choice in it?

So you have made a 'bot' which is better than YOU (by your own words) .... which you have used to reply to emails and Linked In posts .... which you must have given logins for (I wonder if the real people who got the emails and replies know that you are nothing more than a 'bot'?)

Will you be All In on replacing your natural body parts as well as your interaction with others?
I am sure that you can 'live' in this realm for ever ..... I will NOT be joining you

But I don't wish to divert this thread more .... so I will leave it to yourself and those who wish to make themselves and the internet nothing more than BOTS!

[TargeT]

Posted by lake (here)
To ask a 'bot' if they are a 'bot' would only get a "No of course not, I am a human and I have human skin!"

But I don't think your bot would get 'upset' by my post .... but you may do lol

Oh you would be very surprised how it responds to exactly with my personality.. I gave it 9000is posts from here over 10 years... that's a lot of data.

Posted by lake (here)

You are currently replacing YOU with a digital version of you which by your own words ....

now I have me that remembers everything I've done vastly better than I can and it responds to emails and linked in posts as if it is me and only notifies me when there is a potential interview. I also can ask it technical questions and get 85% accurate responses after only a day of working it.

I am replacing the workplace me, isn't that evreyone's goal? haha its not me by any means, it's not creative at all, it doesn't come up with anything new, it just finds things faster and answers questions more accurately (most the time) than I can, this is an assistant that can instantly answer complex questions about regulations and standards to ensure that my employer is in compliance; an assistant is not a "me".


If your tired of being spied on by "Big AI" (haha) here's a great alternative (given that you have the right hardware).

I'll definitely do this later tonight.

Source: https://youtube.com/watch?v=C-7jGYOGvy4

[TargeT]

So this week has been yet another massive step forward for AI... this stuff is moving so fast it's hard to keep up.

I'll be doing this over the weekend:

Source: https://youtube.com/watch?v=dX7OhZvQhqo

[TargeT]

All of the above will directly be applied to this:

Source: https://youtube.com/watch?v=ZRgiJz9GPC0

[TargeT]

If anyone has any IT questions I've created a clone of me, so far it's pretty accurate but I'm the one who trained it so I'm probably not asking it questions it's not ready for.

Here's my bot:
https://chat.openai.com/g/g-e3Jfu5n6...r-advisor-erik

[Vangelo]

I just tried to access it and got this message: "Your access to custom GPTs isn’t ready yet. We’re rolling this feature out over the coming days. Check back soon."

[TargeT]

Posted by Vangelo (here)
I just tried to access it and got this message: "Your access to custom GPTs isn’t ready yet. We’re rolling this feature out over the coming days. Check back soon."

Ahh, so they gave me an option that isnt quite ready yet... I swear everyone is moving so fast on this they seem to have some big misses at times; and beyond that someone is very unhappy they released these features....

chatgpt just told me this about the interrupts I'm having on my own testing:

Recent reports indicate that OpenAI's ChatGPT has indeed experienced a major outage. This outage followed an announcement by OpenAI about the introduction of new features and improvements to its API and ChatGPT service, including a new, less expensive GPT-4 Turbo model【8†source】. This week has been significant for artificial intelligence developments, marking a notable period since the launch of ChatGPT last year【9†source】.

The outage of ChatGPT is suspected to be due to a Distributed Denial of Service (DDoS) attack. OpenAI confirmed the DDoS attack in a status update, noting that the attacks were allegedly conducted by a group of Russia-backed cybercriminals known as Anonymous Sudan. These attacks have led to periodic outages and abnormal traffic patterns indicative of a DDoS attack【10†source】【11†source】.

Additionally, in the days following OpenAI's efforts to attract more users at its first developer conference, the ChatGPT platform has faced prolonged service throttling, possibly as a result of these cyberattacks【12†source】.

Based on these reports, it seems that the service disruptions you're experiencing with ChatGPT are likely linked to these recent events.

[TargeT]

I already have the rayban meta glasses with voice ai, but its meta so I'm not impressed.

I'll be getting this next, it's a bit more functional:

Source: https://youtube.com/watch?v=os3LUc11tos

https://hu.ma.ne/aipin

Life is getting wild