Some thoughts about ChatGPT and AI

Everyone is now talking about this new way of using AI in an interactive form of communication.
When we talk about free or open and AI, these three questions immediately came to my mind:

1. “If you’re not paying for the product, then you are the product.” That’s a quote from Daniel Hövermann in The Social Dilemma. What will be the business model?
2. “Will my Job be replaced?”
3. “Can I trust, and what is my remaining responsibility?”

The blog post is structured in:

• Let’s get started with ChatGPT
• “If you’re not paying for the product, then you are the product.”
• “Will my Job be replaced ?”
• “Can I trust, and what is my remaining responsibility?”
• Summary
• Interesting additional resources

1. Let’s get started with ChatGPT

So, let’s get started with ChatGPT and I was asking ChatGPT some questions about itself.

Here is my initial chat with ChatGPT about ChatGPT.

Here is a quick extract:

• Natural Language Processing is used as the fundamental AI technology.
• Open AI is a private organization that is sponsored.

Let’s take a look at the sponsors, who are they?

Name	Type	Background
Sam Altman	Human	American entrepreneur, investor, and programmer.
Greg Ruckman	Human	Sportsman
Ilva Sutskever	Human	A Canadian computer scientist working in machine learning, who co-founded and serves as Chief Scientist of Open AI.
Wojciech Zaremba	Human	Polish computer scientist, a co-founder of Open AI (2016–now), where he leads both the Codex research and language teams.
Elon Musk	Human	An American business magnate and investor.
Reid Hoffman	Human	An American internet entrepreneur, venture capitalist, podcaster, and author.
Peter Thiel	Human	A German-American billionaire entrepreneur, venture capitalist, and political activist.
Amazon Web Services	Company	An American Cloud provider
Infosys Limited	Company	An Indian multinational information technology company that provides business consulting, information technology, and outsourcing services.

2. “If you don’t pay for a service or product you are the product”

What will be the business model for ChatGPT in the future?

3. Will my Job be replaced ?

AI started out by “replacing boring, tedious, less creative tasks,” so it started analyzing, computing, and researching data.

Now AI started with further creative tasks such as creating digital art (for example artssy), images for example photoleapapp and music (for example unite).

In this context I like the blog post from the beyond.agency Will AI take my job?.

Here is a quick extract of the blog post.

• Jobs that AI will eventually replace, here are some were the risk level over 50%:
  • Customer service executives
  • Bookkeeping and accountants
  • Receptionists
  • Proofreading
  • Manufacturing and pharmaceutical worker
  • Retail services
  • Courier services
  • Taxi and bus drivers

4. “Can I trust, and what is my remaining responsibility?”

When you use AI in a communicative way, like ChatGPT, you get a targeted answer and not a list of choices like a Google search to decide for yourself which result is right for you. This is a way to remove personal responsibility from you. That can be a good thing, because you get results faster that way, but as I said, it’s also a way of evading personal responsibility. This means you need even more that you have to trust that the AI will show you the right results. In other words, who has developed and trained the model and who is responsible for a used foundation model?

The table below shows an assumption to the question: what is your responsibility and most likely will it be replaced more or less in the near future?

Responsibility	Replacement
You only need to remember topics/definitions like written specifications	most likely
You only need to repeat the same tasks.	most likely
You do some kind of creative work based on known facts or examples.	likely
You do some kind of creative work based on unknown facts or examples.	not now

We should need to keep in mind AI for now is still based on models, patterns, calculations, statistics and many more. The question is, what if the data has changed, or maybe regulations, the sense of social justice has changed? Who and how will be the model updated and has the objective of the model changed?

5. Summary

If you now look at ChatGPT it certainly makes sense to look at the data science and AI basis. I think a good starting point is to get a basic understanding of probability calculation, statistics, classification, and the differences between “population data” (is all data available) and “sample data” (are only representative subsets available?) in connection with variance (simplified from my perspective a possible solution area), confidence interval, and error deviations. Surely, there is much more to learn about machine learning and deep learning, neural networks, and many more topics.

With this initial minimal knowledge from my perspective, it becomes clear that, in the end, there are three important points in AI: the models, the data and “what do you want to achieve?”.

These first two points are certainly created to a large extent by humans. The third element is the real driving force and always comes from humans!

In the end, the models are based on assumptions that started at a certain point in time in order to achieve a goal (e.g., We want to write a story! What do you consider to be a good story these days and what are the rules for creating one? What is social? accepted and what not?).

For a short time now, the AI models are still being created by humans for humans and their needs.

When creating a model, you use known rules, laws and relationships that are already known or that are specially developed for a model for a specific need.

The success of AI depends on the trust, laziness and motivation of people who want to achieve their tasks or goals with it in a simplified way. These people have to decide whether they blindly trust the AI and also for more creative activities, because it is faster, easier, and cheaper and they are maybe 80, 90, 95 or even 99 percent satisfied with the solution.

The question remains, is the person using AI really able to assess the result and use it correctly? (e.g., Is the created story good from my point of view? Does it match what I want to express?)

As said: everything depends on models, data and why were they created, i.e. for whom and for what purpose? Are regularities known for the respective subject area or do they have to be found first, e.g., “Literature, what makes a good story?”, With which data was the model trained and tested, who selected the data?

So, for me ChatGPT can and will certainly accelerate creative processes. There is still a need, although fewer, for people to have knowledge of the subject matter for which the models are being made e.g., “How do you write a good story?”.

In the end it remains to be seen for which needs models will be developed by humans for humans in the future and how much trust will be placed in AI without questioning the data and the models.

6. Interesting additional resources

Related blog posts:

• Google An important next step on our AI journey
  • Google introduced Bard which will help to build skills and can be compared to ChatGPT.
  • Bard is using a lightweight model version of the Language Model for Dialogue Applications. (LaMDA) (Pathways Language Model) (MusicLM)
• Microsoft announces surprise event for Tuesday with Bing ChatGPT expected
• IBM: What are foundation models?
• What Is ChatGPT Doing … and Why Does It Work?

YouTube:

• ChatGPT Tutorial - Use ChatGPT for DevOps tasks to 10x Your Productivity by Nana
• IBM Research: Operationalizing Foundation Models | Enterprise AI

---

I hope this was useful to you and let’s see what’s next?

Greetings,

Thomas

#chatgpt, #ai, #trust, #responsibility