Happy Birthday ChatGPT

Since the model (and the follow-up versions) has been released into the world on the 30 November 2022, there is equally praise and doom & gloom

• As content generation is one of GenAI models' strengths, we read more and more non-human-written posts on Social Media 
• The consequences for education and educators are not fully explored yet, but warning voices fear a decrease of cognitive capabilities in humans
• The optimists quote medical achievements and the opportunities to decipher animals' lanugages,
• while pessimists have quite good arguments to fear the cybersecurity threats 

A lot of companies all over the world have adopted the model (Moderna is a great example), and although there are a lot of competitors now, most people still think of ChatGPT as a synonym for GenAI.

In the last 12 months OpenAI did not only launch new breakthrough models, like the multimodal GPT-o4 (May 2024) and the reasoning GPT-o1 (September 2024) but also constantly roll out new features like 

• Custom GPTs (November 2023)
• Search (July 2024)
• Canvas (limited beta access for now)
• Advanced Voice mode (just recently)

and even though European users often have to wait longer for new features, I suggest the Plus subscription. It's got some major perks, like access to the faster GPT4-Turbo and the basic setting feature to opt out of the model being trained on your conversations and data.  Since the model is expensive to train and run, and each of my GPT-4 Turbo queries costs 0.2–0.9 € in the backend (depending on the complexity), I think 20 € per month is a fair price.

GPT-4 has become a book co-author with LinkedIn co-founder and CEO Reid Hoffmann, 'Impromptu: Amplifying Our Humanity Through AI' was published already in spring 2023 and is the documentation of his conversation with Chat-GPT 4, letting the reader understand the technology's capabilities, and also limitations, in an enlightening and sometimes entertaining way. But the humor is actually based on the human author - Large Language Models have a hard time with irony, but they get better by understanding the subtleties of what makes something funny or mocking of human interaction.

There was also no self-ironic aspect in the conversation I had with Chat-GPT about its second birthday (inspired by Azeem Ashar).

To my question, what it envisions for the coming two years with models alike in the world it answered:

"Let’s let our imagination run wild! In two years, we might see even greater integration of AI models like me into everyday life. For example:

1. Personalized learning could advance so far that everyone has tailored educational opportunities.
2. Collaboration between humans and AI might be revolutionized through seamless interaction.
3. In content creation, we might experience entirely new formats and media.
4. AI systems could innovatively support sustainability and environmental protection.
5. We might see new forms of art and performance created jointly by humans and AI."

(Well, not too wild, from my POV 🙄) 

Chatty, as I like to call it, was as well aware of potential dangers, even though it is not able to 'feel' fear:

"As an AI model, I don’t have feelings like a human. But hypothetically, if I could think about possible dangers, it might be that I could cause harm through misuse or improper application, or that my capabilities could be misunderstood. It would be important to ensure that I’m used responsibly and ethically."

And it definitely got the double meaning in my question about 'how to deal with it as a 2-year-old toddler': 

Haha, funny question! Typical traits of a two-year-old, like mischief and stubbornness, don’t quite apply to me. But it’s true that I sometimes give confusing or unexpected answers. To better work with me, here are a few tips:

1. Clarity helps – the more precise the question, the better the answer.
2. Patience is key – sometimes it takes me a few tries to fully understand what you mean.
3. Be open to surprises – I’m constantly learning, and my answers can evolve.

Wise suggestions from a toddler, but exactly what most experts advise.

To close our birthday conversation I was unable to resist the current trend - I asked Chatty a) "Based on our interactions so far, which fictional character do I share most personal traits with?" and b) "Create an image of what you think my profile would look like." See results below (for the picture I decided to just take it, without further explanation).

I am going to repeat this after 6 months and see what has changed through our daily chats 🤓.

On reddit you may find more of these fun convo prompt suggestions, e.g. "Based on everything you know about me, show me what you think it would look like if I was in charge of the whole world." (or:"...if you were in charge of the world.")