Maybe I’m late to the party, I don’t know, but it took me some time to process it. I gave it a lot of thought and I’m ready to share my insights with you.

Oh… sorry… you have no idea what I’m talking about. 

Ok, so today’s post is gonna be a bit different, and a bit more philosophical in nature, rather than the pragmatic advice I usually provide in my posts. I am going to talk about ChatGPT and the future of AI as I see it.

Even my mother already heard about ChatGPT (and not from me), so I assume that if you are reading this post it means that you are connected to the Internet and hence heard about it as well, or better – maybe even tried it (and if not – what are you waiting for? For now it’s free to try).

Now, I’m not sure if all of you know, but as of this time of writing, aside from being a product leader, I’m also fulfilling another role and that’s being the head of data science of my business unit. My team deals a lot with NLP (Natural Language Processing) as part of their day-2-day job – so I think I have some meaningful perspective on this topic.

 

ChatGPT

Let’s start from this specific product and then we’ll drift away to discuss the future of AI in general.

So here is the thing – as part of my job – I’ve been exposed to many NLP frameworks and solutions out there. I played with many of them and evaluated their results. Until I tried ChatGPT the quality of the results provided were on the scale from ‘very bad’ to ‘cute’.

Until ChatGPT I considered Google to have the most advanced NLP technology out there, given what I experienced by using their search engine and their Google News product.

ChatGPT changed all of that.

You see – most technologies out there are not really disruptive. Even the unicorn companies of the last few years – AirBnb, Uber, Spotify, Monday.com, Lemonade and all the others – haven’t introduced anything which is truly disruptive. Their technology is decent, but their main achievement is that they managed to provide a product that solved a real problem for a real market – and this is why they became so successful (and a big kudos for them on that).

That being said, if I was asked to clone their product and was given a team of decent engineers I definitely believe it’d be feasible to build such a clone. In fact, most individuals who have a strong engineering background can probably reverse-engineer the technology behind each of these companies without too much of a hassle (or at least have a strong understanding of how it was built).

However, as I see it, ChatGPT is truly disruptive. It’s demonstrating a strong technology that is a game changer to everything that was before. If you provide me with a group of talented data scientists and engineers – I am no longer sure I can deliver a product in a similar quality within a reasonable amount of time (I guess it will take at least a couple of years and even then I’m not sure I will be able to come close to the same quality).

 

This technology beats all existing NLP products I know in a knock-out and it’s better than everything I’ve seen before by several orders of magnitude.

Some say it’s the most disruptive thing since the invention of the Internet. I’d suffice by saying it’s the first disruptive product I have seen since the announcement of the iPhone in 2007.

 

For those who haven’t tried the product themselves or not sure they understand why it’s unique – let’s mention some of its capabilities:

1. It can answer intelligently to most questions which are based on factual information. For example: can you rank George Clooney movies according to their success in the theater?  
2. It can provide an intelligent opinion for endless topics. For example: For someone who is going to visit Spain for the first time – would you recommend visiting Madrid or Barcelona and why? How would you  suggest checking whether Atlantis ever existed or not?
3. It can generate text for almost any purpose (and this is its most amazing capability). For example:

Can you suggest a blog post about product management and roadmap planning?

[The bot provides me with an outline of this post. I wanted the full post, so I’m clarifying my request:]

Can you write this full post?

[The answer to this is in the appendix below. It wrote the post immediately, without any delays]

It can also write code!! For example: 

In Java we defined a class called Person with the members ‘name’ and ‘address’. Can you suggest a method that receives a list of Persons and return true whether the current instance is included in the list?

[The bot provides this method immediately + an example of how to use it. I challenge it a bit further:]

Can you add a method to this class that given a list of Persons return all Persons who live on the same address as the current instance?

[The bot delivers the code immediately + an example again]

 

I know some people claim that the answers are quite basic and the ‘creative’ capabilities of this bot are following well known practices. Although it may be so for some of the questions and some of the tasks assigned to it – I think these people are missing the point:

This is just the first version known to the public. Give it a few more years and I am quite positive that at least when it comes to questions being asked – it’d be hard to tell whether a human or a bot provided the answers.

When most of today’s NLP products are still struggling with proper entity extraction (what ‘entities’ are discussed within each post) this product displays full comprehension of what it’s being asked to do, how to gather the proper information from its training set and provides a fully readable answer that actually makes a lot of sense and it’s accurate.

I’m truly amazed by this product. 

 

What does it mean for the future of AI?

Alan Turing came up with the Turing Test in 1950 for determining whether a machine can demonstrate true intelligence. You can read the link I provided for a full description of the test, but in short the essence of this test is for a human to ask two participants, who are not present in the same room and can be communicated only via text, any question they want. Now, one of the participants is a human, and the other one is a machine.

If the human who is asking the questions can’t tell at the end of the questionnaire which of the participants is a human and which is a machine – then the machine has ‘passed’ the test and according to Turing it possesses true intelligence.

I truly believe that we are very near this point in time, if we haven’t passed it already, that the machines can successfully pass the Turing Test. ChatGPT has definitely raised the bar on that aspect.

But what does it really mean? When I studied computer science in the late 90s – the Turing Test seemed like a decent test for machine’s intelligence. The existing capabilities were so far from the required threshold that everyone could agree that this is a fair test for machine’s intelligence.

Fast forwarding to today, now that we’re very close to machines being able to pass this test – I no longer believe this test is good enough.

This is where this post is becoming more philosophical in nature, so pardon me in advance.

They say that ‘if it walks like a duck, sounds like a duck and looks like a duck – it’s a duck’.

So we already have robots who can walk like humans, we can definitely make them sound like humans and if we integrate ChatGPT into their ‘brain’ – they can almost talk like humans.

So in that case – is the robot a human? Or is it similar to humans?

 

Let me ask you the same question from a different angle:

If you have a machine that beats the world class champion in chess – does it know how to play chess?

As for the second question – I actually had this debate with a friend in 1998 while walking in the corridors of the faculty. 

 

My answer to both questions is ‘no’. 

Even though robots can look like humans and demonstrate behavior patterns which are similar to humans – they are still not humans.

And ‘no’ – a machine which wins everyone in chess doesn’t really know how to play chess IMHO.

 

I know there are a lot of people who believe I’m wrong and disagree with me. That’s totally fine.

If you want to understand why this is my answer there is no other way other than to dive into the concepts of consciousness and the human emotions & fears.

I strongly believe that no matter how advanced our AI is going to be – we won’t be able to infuse a true consciousness into a machine. Additionally, a machine will never be able to ‘feel’ or truly ‘fear’.

We may ‘teach’ it to demonstrate feelings and fears (heck… we’re doing it in computer games already), but it’s just following some instructions that are written in a program.

I think the true achievement of AI in the last years was eliminating the ‘deterministic’ behavior of the machine, and designing a software that can expand itself automatically and evolve.

Isn’t it similar to life? On some aspects yes, but on many other aspects it’s not enough because the core is still ‘programmed’. 

That being said – even without having a true consciousness, the AI can be evolved to produce some incredible stuff.

For example, they say that within 3-5 years – computers will be able to produce songs. To me that sounds very feasible and I find this both amazing and troubling.

 

In the short term – it’s gonna make everyone more lazy. I already read that some writers are using ChatGPT to produce the initial (or full) text of their stories. I never gonna do it, but I do understand why someone is being paid by the hour to write content might want to do this. The end result is even more ‘clickbait’ articles spamming the web (5 things you must do after breaking up…).

I guess that we’ll see the same when it comes to music & songs. Yes, machines will be able to produce songs where the melody and words make sense, and might even be good for dancing. However, it’s just gonna follow existing patterns, and hence it’s all going to sound the same.

 

Looking on the longer term – will machines be able to produce creations which are original and have a genius spark in them? I don’t think so. Maybe only by coincidence.

I believe that the ‘genius spark’ will be kept to humans alone, though given the dumbness our world will be driven into (really… everything is being done by machines) – we might lose our touch with our creative nature.

 

Do I believe machines can take over the world like in the Terminator movie series?

Sadly yes…

Regardless of my belief that machines will never have any true consciousness, we can develop a ‘smart’ program that optimizes for the extinction of the human race. Actually, and very sadly, I’m not sure we’re that far from such a capability. All we need is a talented ‘evil’ programmer and machine access to nuclear weapon systems.

Very sadly, again, this is not far fetched anymore.

But I don’t want to end this post with a gloomy tone. 🙂

 

I do see a path, which requires discipline and intent, where we can evolve the existing state of AI to something incredible that actually serves the human race. With proper leadership, we can leverage AI to optimize for the survival of the human race rather than its extinction. Such optimizations, for example, will include automatic calculations and decisions of what fields to irrigate and what to plant on any field in the world so there will be enough food supply for everyone. The AI can also assist governments with proper population growth strategies to make sure countries will never get over-populated.

Those are just a few examples. If AI is put into good use – it can really make our life better and actually eliminate the ‘survival’ aspect of our lives so we can finally find the time to connect to our creative nature 🙂

 

So… that’s it for today!

If you found this post/series useful – let me know in the comments. If you think others can benefit from it – feel free to share it with them.

 

Thank you, and until next time 🙂

 

Appendix:

Certainly! Here is a full blog post on product management and roadmap planning:

Product Roadmap Planning: A Step-by-Step Guide

As a product manager, it’s your job to guide the development and evolution of your product. One of the most important tools you have at your disposal is the product roadmap. A product roadmap is a high-level visual summary that outlines the direction and focus of your product over a specific period of time. It provides a clear, actionable plan for your team to follow and helps align stakeholders around a shared vision.

But creating a product roadmap isn’t always easy. It requires careful planning, collaboration, and the ability to balance short-term and long-term goals. In this post, we’ll walk you through the process of creating a product roadmap, step by step.

Step 1: Understand the purpose of a product roadmap

Before you begin creating your roadmap, it’s important to understand the purpose it serves. A product roadmap is not a detailed project plan or a list of all the features you want to build. Instead, it’s a high-level overview of the most important goals and objectives for your product. It should answer the following questions:

• What is the overall vision for the product?
• What are the key features or initiatives that will help us achieve this vision?
• When do we expect to complete these features or initiatives?

Your product roadmap should be flexible enough to allow for changes and adjustments as the market and customer needs evolve.

Step 2: Gather input from stakeholders

To create a product roadmap that aligns with the needs and goals of your organization, it’s essential to gather input from key stakeholders. This includes your team, your customers, and any other departments or teams that have a vested interest in the success of your product.

Some questions to consider when gathering input from stakeholders include:

• What are the top priorities for the product?
• What challenges or opportunities do we need to address in the short term?
• What long-term goals do we need to consider?

Step 3: Define the scope of your roadmap

After gathering input from stakeholders, the next step is to define the scope of your roadmap. This includes determining the time frame for your roadmap and the level of detail it should include.

Most product roadmaps cover a period of 6-12 months, but it’s important to consider the unique needs and goals of your product when determining the right time frame. For example, if you have a product with a shorter development cycle, you may want to create a roadmap with a shorter time frame.

Step 4: Identify and prioritize features

Once you have a clear understanding of your goals and the time frame for your roadmap, it’s time to start identifying and prioritizing the features and initiatives you want to include.

To prioritize your features, consider the following:

• How does each feature support the overall vision for the product?
• What is the potential impact of each feature on the product and the business?
• How feasible is it to develop and implement each feature within the given time frame?

Step 5: Estimate development time and resources

Once you have a list of prioritized features, the next step is to estimate the development time and resources required for each one. This will help you understand what is realistically achievable within the time frame of your roadmap.

Step 6: Create a timeline for release

With your prioritized list of features and estimates for development time and resources, you can start creating a timeline for release. This timeline should include key milestones and dates for each feature.

Step 7: Communicate the roadmap to stakeholders