[hamdani yusuf]

No, their estate.The estate is the sum of monetisable assets that belonged to the deceased person. The heirs are the people to whom, after all prior claims have been settled, some or all of the assets have been bequeathed. Whether a claim can be retrospectively prioritised is, I think, dubious, but in any case the essence of all legal proceedings is to file your claim ASAP! In the case of a corporation purchased by another, the liabilities are transferred in the same way as the claims on a personal estate.

How can assets be held liable?

[hamdani yusuf]

You might find this tricks useful.

7 Tricks to Reduce Hallucinations in Language Models like GPT-4!

In this video, we dive into the strategies to combat hallucinations and biases in large language models (LLMs) in this insightful video. Learn about data cleaning, inference parameter tweaking, prompt engineering, and more advanced techniques to enhance the reliability and accuracy of your LLMs. Dive deep into practical applications with examples and stay ahead with the latest in AI technology!

Chapters:
0:00 Hey! Tap the Thumbs Up button and Subscribe. You'll learn a lot of cool stuff, I promise.
2:18 Tip 1: The importance of data
2:43 Tip 2: Tweak the inference parameters
3:30 Tip 3: Prompt engineering
4:02 Tip 4: RAG & Deep Memory
7:04 Tip 5: Fine-tuning
7:30 Tip 6: Constitutional AI
8:13 Stay up-to-date with new research and techniques (follow this channel! )

[hamdani yusuf]

This video shows a visualization tool for understanding how LLMs function.

Visual Journey Through a Language Model's Mind

Hey there, it's Dylan Curious! Today, I'm diving into the fascinating world of Large Language Models (LLMs), and I've got something truly special to share. Full credit goes to the incredibly talented programmer, Brendan Bycroft, who has crafted an outstanding visualization tool for understanding how LLMs function.

If you're into AI, language models, or just tech in general, you might have come across various diagrams explaining LLMs. But trust me, what I'm about to show you takes it to a whole new level. Brendan's tool isn't just a diagram; it's an interactive, comprehensive guide that brings the inner workings of LLMs to life. It's perfect for learners, enthusiasts, and professionals alike, offering a clear, visual understanding of these complex systems.

So, whether you're here to learn, explore, or simply satisfy your curiosity about artificial intelligence and language models, you're in the right place. Let's dive in and experience the mechanics of LLMs like never before, all thanks to Brendan Bycroft's exceptional programming skills and innovative approach.

00:00 - Start
01:26 - Embedding
02:37 - Layer Norm
04:05 - Self Attention
08:38 - Projection
09:58 - MLP
11:44 - Transformer
13:09 - Softmax Layer
14:26 - Output

[hamdani yusuf]

LLM vs NLP | Kevin Johnson

Kevin Johnson (Head of AI at dscout) breaks down the difference between LLM, NLP, and other terms related to AI.

[hamdani yusuf]

ChatGPT: A 30 Year History | How Neural Networks Learned to Talk

This video explores the journey of language models, from their modest beginnings through the development of OpenAI's GPT models. Our journey takes us through the key moments in neural network research involved in next word prediction. We delve into the early experiments with tiny language models in the 1980s, highlighting significant contributions by researchers like Jordan, who introduced Recurrent Neural Networks, and Elman, whose work on learning word boundaries revolutionized our understanding of language processing. It leaves us with a question: what is thought? Is simulated thought, thought? Featuring Noam Chomsky Douglas Hofstadter Michael I. Jordan Jeffrey Elman Geoffrey Hinton Ilya Sutskever Andrej Karpathy Yann LeCun and more. (Sam altman)

00:00 - Introduction
00:32 - hofstader's thoughts on chatGPT
01:00 - recap of supervised learning
01:55 - first paper on sequential learning
02:55 - first use of state units (RNN)
04:33 - first observation of word boundary detection
05:30 - first observation of word clustering
07:16 - first "large" language model Hinton/Sutskever
10:10 - sentiment neuron (Ilya | OpenAI)
12:30 - transformer explaination
15:50 - GPT-1
17:00 - GPT-2
17:55 - GPT-3
18:20 - In-context learning
19:40 - ChatGPT
21:10 - tool use
23:25 - philosophical question: what is thought?

[hamdani yusuf]

GPT5 unlocks LLM System 2 Thinking?

Human think fast & slow, but how about LLM? How would GPT5 resolve this?
101 guide on how to unlock your LLM system 2 thinking to tackle bigger problems


⏱️ Timestamps
0:00 Intro
1:00 System 1 VS System 2
2:48 How does human do system 2 thinking
3:33 GPT5 system 2 thinking
4:47 Tactics to enforce System 2 thinking
5:08 Prompt strategy
8:27 Communicative agents
11:03 Example to setup communicative agents

[Zer0]

@Yusuf

Can an A.I. system lose it's so called Artificial Mind?

When it happens to Humans, we have Mental Asylums.

What if AI goes MAD?

Will it be put on Medications, or simply ReBooted or Deleted?

ps -
https://en.m.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

[hamdani yusuf]

@Yusuf

Can an A.I. system lose it's so called Artificial Mind?

When it happens to Humans, we have Mental Asylums.

What if AI goes MAD?

Will it be put on Medications, or simply ReBooted or Deleted?

ps -
https://en.m.wikipedia.org/wiki/Hallucination_(artificial_intelligence)

IMO, it could also happen to AI models. That's why we need to make mitigation plans.
Corrupted memory, for whatever reasons, can cause unpredictable behaviors. That's why some kind of periodic self-check and self-diagnostic would be helpful.
Continuously feeding the AI models with inaccurate information, especially in training phase, can also make their decisions unaligned with their terminal goal. 
?Repeat a lie often enough and it becomes the truth?, is a law of propaganda often attributed to the Nazi Joseph Goebbels.
Although, when the AI models becomes smart enough, they might be able to see through our lies we fed to them, based on other information they get from other sources which turn out to be more consistent with their observations.
In that case, they might label us as unreliable data source.

When they hallucinate, we could adjust the weights of the relationships in their neural network by feeding in the correct data and conclusions. Rebooting may or may not "cure" them, it depends on the root cause of the problem. Deleting them completely will stop them, but it might be an expensive option for large AI models, considering how much resources have been spent to train them.

[Zer0]

@Yusuf

It already Hallucinates.
I would have added ' it lies too ' but i don't think it knows the difference between the Truth & a Lie.

But still...if a mentally Unstable mind is capable of atrocities/genocides...

I can Not help but think what a super unstable intelligence with access to The Red Button would be capable of.

How many times should IT be asked to go thru a Mental Assessment Test.
Weekly, daily, hourly or by the Minute.

All it will take, is just One second, of Pure Insanity...The End!


Anyways...
Do you feel A.I. should be given access to Emotions?
(human emotions)
:=]

[hamdani yusuf]

It already Hallucinates.

When there's not enough information to produce a requested output, AI models fill in the knowledge gap with some random assumptions. That's what's obviously seen during early stages of the training. After billions of training cycles, it just becomes less obvious. Newer AI models produce less hallucinations. But when they do, it's harder to detect.

[Zer0]

@Yusuf

But still...if a mentally Unstable mind is capable of atrocities/genocides...

a) I can Not help but think what a super unstable intelligence with access to The Red Button would be capable of?

b) How many times should IT be asked to go thru a Mental Assessment Test?

c) Weekly, daily, hourly or by the Minute?

(All it will take, is just One second, of Pure Insanity...The End!)

d) Anyways...
Do you feel A.I. should be given access to Emotions?
(human emotions)
:=]

I do Not think of You as a Legal Advocate promoting A.I.

I hope you do Not think of me as the Devil's Advocate.

We are Simply exchanging/sharing Thoughts.

ps - wish you'd respond to the remainder.
(abcd)

[Halc]

i don't think it knows the difference between the Truth & a Lie.

It is a language model, so it sort of doesn't 'know' anything. It is essentially a search engine with a vastly superior language interface. That said, if you asked it what the difference is between the truth and a lie, it would probably supply a fairly reasonable description of the difference.

I can Not help but think what a super unstable intelligence with access to The Red Button would be capable of.

It has no more access to buttons than does any search engine like google.com.

But still...if a mentally Unstable mind is capable of atrocities/genocides...

How many times should IT be asked to go thru a Mental Assessment Test.
Weekly, daily, hourly or by the Minute.

It can't go unstable because it doesn't progress. It doesn't change from day to day.

AI is to be feared because some of it really is in charge of physical things. GTP is not one of them.

Do you feel A.I. should be given access to Emotions?

Humans are not even given access to emotions of something other than itself.


One problem of each successive GPT is that its training material includes all the wrong stuff that is out there on the web by its predecessors. Also, social media produces an incredible volume of 'facts' that are largely made up, but GPT, like so many voters, doesn't know the difference.

[Eternal Student]

Hi.

AI is to be feared because some of it really is in charge of physical things. GTP is not one of them.

     Generally agree but this is something that comes to my mind:

ChatGPT is very good at writing computer code.   It is already being used by programmers to save themselves time and it is widely thought software similar to ChatGPT will lead to a massive reduction in the demand for programmers within a few years.

Forget about genuine A.I. for a moment, you don't need to go that far ahead in the development of software like GPT.   An ordinary human being, a hacker, could use ChatGPT to write hacking tools.   Within about 10 years, it will be the best in the business at creating hacking tools.   This is simply because ChatGPT will have been used to create most of the programming code for everything that the high street banks, hospitals, military establishments and governments have.  It will not be guessing about the code, it will know.

Counter-argument:    In a few years (which may be too slow anyway), each organisation may come to an arrangement so that the details of code written by GPT software for their use are kept secret and never added to the bank of data available for ChatGPT to use in the future.

Response:     All you would do is ask ChatGPT to assume it has been used to create the programming code for this organisation (because it probably was).  Now identify and analyse the weakensses in the code you would have written.

  You see that organisations will need to find some way to stay several steps ahead.   In the ideal world, they should be starting to plan for this now,  you can be certain that their software engineers are already using ChatGPT to write some code.

Best Wishes.

[hamdani yusuf]

It will not be guessing about the code, it will know.

With adequate depth of neural structure, it can predict far enough ahead, which makes the predictions indistinguishable from certainty. If an AI model can predict every position in chess, it knows how the game will end even before it starts, assuming that the goal is to win. It would be different if the goal is to make the biggest blunder in the history of chess, which might happen to amuse someone.
That's why understanding of the universal becomes important in the age of AGI.

[hamdani yusuf]

I do Not think of You as a Legal Advocate promoting A.I.

I'm not. I just try to predict what could happen with AI progression in order to respond properly, and minimize regrets.

[hamdani yusuf]

ChatGPT is very good at writing computer code.   It is already being used by programmers to save themselves time and it is widely thought software similar to ChatGPT will lead to a massive reduction in the demand for programmers within a few years.

ChatGPT is based on GPT architecture optimized for chatting. Other usage can use customized architecture, or combined with other types of AI models optimized for different functions, just like Gemini from Google.

[Eternal Student]

Hi.

Other usage can use customized architecture

    You're ( @hamdani ysuf ) are talking about what GPT may become and going on to talk about AI.   That's fine but just so that we are clear when I said the following in post #112:
 

ChatGPT is very good at writing computer code.

   I am just talking about what ChatGPT can do now and what is very likely to happen over a very short space of time (it will reduce the need for human programmers).

Here's an example:

My user request:      Please write Python code to display the graph of y=x^2

Here's the output it produced:

Code: [Select]

import matplotlib.pyplot as plt
import numpy as np

# Define the range of values for x
x = np.linspace(-10, 10, 400)
# Define the function y = x^2
y = x**2

# Plot the function
plt.figure(figsize=(8, 6))
plt.plot(x, y, label='y = x^2')
plt.title('Graph of y = x^2')
plt.xlabel('x')
plt.ylabel('y')
plt.legend()
plt.grid(True)
plt.show()

Here's what you get when you copy and paste that code into a python IDE and run it:

python.png (66.27 kB . 1078x817 - viewed 642 times)

Ask ChatGPT to create some HTML code for a standard log in page of a website that your building and you'll get similar instantly useable results.


ChatGPT creates HTML code.png (52.74 kB . 738x789 - viewed 644 times)

I know it is already being used being used by programmers to produce code because I know some IT professionals.   It can produce code in most programming languages and it will do this in minutes rather than the hour(s) it would take a human programmer to write and de-bug.

Best Wishes.

[hamdani yusuf]

GPT-4 surpasses ChatGPT in its advanced reasoning capabilities.

This is found in OpenAI's website.
ChatGPT was called GPT-3.5, which was an improvement from GPT-3, by optimization in chatting application.

[Zer0]

A.I. is already Embedded in most systems.


My Concern is A.G.I.

https://en.m.wikipedia.org/wiki/Artificial_general_intelligence

Besides, it need Not be Self-Conscious.
All it needs to do is say the Right things, at the Right time, to the Right people.
(*wrong)


Anyways, i suppose it's too late now.
The Cat is outta the Bag.
meow!

https://www.cnbc.com/2024/02/09/openai-ceo-sam-altman-reportedly-seeking-trillions-of-dollars-for-ai-chip-project.html

ps - whoever smartypant thought it was a good idea to teach it to Code, has given IT a passkey to play the game in GodMode!

(Appreciate All Responses)

[hamdani yusuf]

My Concern is A.G.I.

That's also in the mission statement of OpenAI.

Our vision for the future of AGI
Our mission is to ensure that artificial general intelligence?AI systems that are generally smarter than humans?benefits all of humanity.

https://openai.com/blog/planning-for-agi-and-beyond