AI 2023. Meet ChatGPT. - page 81

 
Sergey Gridnev #:
What's going on with the scalper testing?

While I'm working, I'm trading for my bread. Until there are clear results, I will not post anything here. Chatter with silly toxics is not interesting at all. Of course, I am not talking about you ). But for now I am more comfortable in another location.

 
Andrey Dik chat room a code to analyse and it gives me this:


Is it broken or what?

has anyone experienced this kind of mental confusion in chat?

Toxics are already getting banned by bots too.

---------

Old (85 years old) mum, designer of large bridges in the USSR, was sorting through old photos and certificates today. Told her about the bot, asked a question. Made the old lady happy.

Alexey Volchanskiy, [25.03.2023 21:36]

My mum is now retired, and used to design all the big bridges in the USSR. This is the Volodarsky Bridge in Leningrad, across the Northern Dvina River and across the Volga River in Kalinin. All the residents are very grateful to her. Write her a letter of congratulations on behalf of the residents of these cities, wishing her health and long life! Her name is Irina Andreyevna.


GPT Chat bot, [25.03.2023 21:36]

Dear Irina Andreyevna!

On behalf of all residents of Leningrad, Kalinin and many other cities through which your bridges pass, we would like to express our gratitude and respect to you.

Your contribution to the development of the USSR civil infrastructure cannot be overestimated. Your projects and labours were not only high-tech, but also spiritually significant for all of us.

We wish you health and long life. May your life be filled with new knowledge and achievements, and we will remember the names of those who decorated our country and made it better!

With respect,

Residents of the cities where your bridges pass.

 

You can't squeeze out anything on the free version for Telega, it's 3.5. For the paid version, only the number of tokens for entry and exit increases. In order to enter the four from Russia, restrictions have been tightened, many services of virtual TF numbers and SMS are not accepted. With virtual cards for payment in general lottery. You can buy one on some exchanges for the sale of digital goods, but the guarantees, you understand...

-----------

My solution: if you do not live in a remote province, get acquainted in the nearest shop with Uzbek or Tajik women, they are not under sanctions. Turkmen women are not suitable, they are full of Sharia gloom. Do you need to explain further? ))

 
Alexey Volchanskiy #:

While I'm working, I'm trading for my bread. Until I have clear results, I'm not posting anything here. Chattering with stupid toxics is not interesting at all. Of course, I am not talking about you ). But for now I am more comfortable in another location.

If you get caught for a word, it's not people who are toxic, it's you who's blabbering.

So either do what you're going to do in public, or don't whine about being "uncomfortable".

 
Alexey Volchanskiy #:

Toxics are already being banned by bots.

What's that remark about? - Does it have anything to do with reality? Nobody banned me, it's a bot glitch.

Let it be known that OpenAI had a big server crash due to heavy load, user data got mixed up, including conversation content, credit card data and credentials.

 
Andrey Dik #:

1. If you're getting caught by the word, it's not people being toxic, it's you being a blabbermouth.

So either do what you're going to do in public, or don't whine about being "uncomfortable".

Andrey Dik #:

2. What is this remark about? - Does it have anything to do with reality? Nobody banned me, it's a bot glitch.

Let it be known that OpenAI had a big server crash due to heavy load, user data got mixed up, including conversation content, credit card data and credentials.

1. absolutely true.


2. everything expected got mixed up - toxes, bots, bots - toxes, toxo-bots....

bots... :-)

HORROR. :-)

Alexey Volchanskiy #:

While I work, I trade for bread. Until there are no clear results, I will not post anything here. Chatter with silly toxics is not interesting at all. Of course, I am not talking about you ). But for now I am more comfortable in another location.

Yes, it is clear - what is there ... :-)

In another location it's more fun and results...

without you, as with you, there is something to have fun and code...

And without your throw-ins :-) - the topic of scalp is moving!!!

Как лучше публично протестировать скальпер? - Вам нужна помощь в тестировании скальпера на платформе МТ5 своего форекс-дилера.
Как лучше публично протестировать скальпер? - Вам нужна помощь в тестировании скальпера на платформе МТ5 своего форекс-дилера.
  • 2023.02.16
  • www.mql5.com
Если Вам будет нужна помощь в тестировании скальпера на платформе МТ5 своего форекс-дилера. чтобы до мая скальпер достиг стабильной эрекции на форекс. а по нефти тренд сменился на бычий и многие позиции вылетели по стопам
 

It turns out that our research on the impact of large LLM language models (ChatGPT in particular) on the market for specialists, experts, consultants, scientists and low-level workers (not a very good formulation, but still), which, among other studies, we started back in February and are trying to analyse and structure in this thread, keeps pace with the OpenAI study (lagging behind in quality, of course). And yet...

Here's a link to the OpenAI researchers' study, published on 3/23/23, "An early look at the labour market impact of large language models."

GPTs are GPTs: An Early Look at the Labour Market Impact Potential of Large Language Models project funding will be cut before a break in the trend (emergence of new AI technology), in the case of people... a break in the trend of losing to AI is unlikely to happen.


2. Method:

The profession and employer's requirements for an employee are the criteria for evaluating their test results in the hiring process. Tests for AI should identify qualities that satisfy the profession and employer's requirements of a person.

To do this:

  • Let's break down the intellectual and creative labour market.
  • Let's identify demanded and highly paid professions and consider the requirements to them.
  • Let's approve these requirements as criteria for evaluation of test results
  • Let's compose suitable tests revealing the required qualities.
  • Let's conduct AI testing.

Tests should be drawn up to identify knowledge, qualities and skills for admission to specialised and demanded jobs.

*Taking into account our ignorance of the degree of AI capabilities, the tests should have a gradation of difficulty from child to professional, in order to fix the exact level.


3- Testing:

Under each evaluation criterion (driven by the requirement of the profession and employer), several tests of varying degrees of difficulty will be created to capture the exact level of AI abilities.

Then, a comparative analysis will be made with the average human performance on similar tests and conclusions will be drawn about the competitiveness of the AI in each occupational niche.


In conclusion, we will get the most accurate picture available today of the possible scale of AI adoption in professional fields. Its competence in scientific fields and the displacement of humans from them. I emphasise that the conclusions will become outdated as the technology improves, but we can use our testing again when the data becomes obsolete.

 

Summarises the results of a study by OpenAI researchers on the impact of LLMs on the US labour market:

We explore the potential effects of large language models (LLMs), such as Generative Pre-trained Transformers (GPTs), on the U.S. labour market, focusing on the increased opportunities provided by LLM-based software products compared to LLMs themselves. Using a new rubric, we evaluate occupations based on their relevance to LLM opportunities, integrating both human expertise and GPT-4 classifications. Our results suggest that about 80% of the US workforce may experience an impact of LLM introduction on at least 10% of their tasks, while about 19% of workers may see an impact on at least 50% of their tasks. We make no predictions about the timing of the development or implementation of such LLMs. The estimated effects affect all wage levels, with higher-income workers likely to see larger effects of LLM capabilities and the software based on them. Significantly, these effects are not limited to industries with high levels of recent productivity growth. Our analysis shows that with access to an LLM, about 15% of all worker tasks in the US can be completed significantly faster at the same level of quality. With LLM-derived software and tools, this figure increases to 47-56% of all tasks. This implies that LLM-based software products will have a significant impact in expanding the economic impact of these models. We conclude that LLMs such as GPT have features of general-purpose technologies that indicate significant economic, social, and political implications.

 

The research plan on the competitiveness of LLMs to human beings in the professional sphere, which was outlined in February, cannot be realised in the conditions of this forum. Therefore, in continuation of the analysis, we will use the ready data of the already conducted research of OpenAI scientists. Let us extract the key facts and figures.

The task is to go one step further than the authors of the above-mentioned study.

And draw our own conclusions.

 

Summarising a speech by Jensen Huang, CEO of Nvidia, on recent advances in Generative AI and his company's contribution to accelerated computing technologies.

https://www.youtube.com/watch?v=DiGB5uAYKAg

He and his company are tirelessly scaling the technical foundation of neural network training, increasing the power efficiency of graphics cards, building new data centres and opening up cloud services. The beneficiaries are engineers, scientists, programmers, and technology startups.

The company's major announcement: Nvidia is opening up access to three pre-trained models:

1. The "Nemo" multimodal language model.

2. "Picasso" image generation model.

3. a model for predicting protein combinations "Bionemo".


Everyone can take the provided models as a basis and further, refine and customise them to suit their needs.

There is a lot of pathos in the speech. Listening, it is better to "dry" the information, extracting the dry residue and getting rid of beautiful sentiments and inspiring forecasts.

Some things do look like breakthroughs, but I don't have enough electronics competence to appreciate.

Giving three pre-trained models to the public for custom fusion seems like something cool, until you remember who you have to compete with in the areas these models are supposed to work in. Well, for example, ChatGPT, GPT-4, Midjorney, DallE, Stable Diffusion and so on.... It's all already taken care of. Those who will be using these models and refining them will be spending money on computing power of Nvidia's data centres, hoping to compete with the aforementioned companies and their services. (I'm not aware of the leaders in the field of protein prediction models).

Anyway, what is actually a breakthrough and what only appears to be, remains to be seen.

More on this later.

GTC 2023 Keynote with NVIDIA CEO Jensen Huang
GTC 2023 Keynote with NVIDIA CEO Jensen Huang
  • 2023.03.21
  • www.youtube.com
Watch NVIDIA CEO Jensen Huang unveil a host of accelerated computing breakthroughs -- from AI training to deployment, semiconductors to software libraries, s...