Predictions 2023: ChatGPT shakes everything up!

It’s that time of year again when we look back at what has technically happened this year and what the predictions are for next year. 2022 was certainly an interesting year that will make the predictions for the next year much less uncertain, let’s review them.

Virtual reality (vr) and the metaverse seem to have entered a winter despite Meta really investing heavily in vr. The ‘launch event’, which was supposed to give an insight into the metaverse, was a cringe-fest with ditto price drops, and six people turned up in an EU VR event that cost almost four hundred euros. I don’t see how this trend can be reversed in the short term, except for small advances for a minority of enthusiasts like me.

However, winter still sounds like a party if you’re into crypto. Cryptocurrency and everything around it has ended in an ice age just to avoid the English word ‘mass extinction event’. Stagnating is one thing, but now pretty much everything in the industry looks real ponzi schemes there are also no appealing applications and no problems are solved. The word blockchain is completely meaningless, but I’ll get back to that in a moment.

The prevailing thought in the industry is that governments and centralized agencies cause inflation and keep printing money endlessly, but now it turns out that the cryptocurrency industry is doing the same, and has no business being on the other side of the balance sheet. After all, we are in a crisis, and a crisis hurts, which among other things results in inflation. There are no elegant solutions, and cryptocurrency is certainly not a solution. It has become clear this year without me needing to give examples. Well, one saw: web3 as the overall principle behind NFTs seems to fly into the ice age even faster than the cryptocurrency itself. Then we are good from there.

Trough of disillusionment?

“It’s not even possible to run trains unmanned”

Now there were also highlights. The launch and success of the James Webb Space Telescope and a more or less successful attempt to generate nuclear fusion that provides more energy than it requires. Now it only seems that we should not expect anything from nuclear fusion in the coming years, while the development of solar and wind is going so fast that they really have a future. The breakthrough we are now hoping for is how we can store this energy. Energy storage in any form now seems to be the big thing we’ve been eagerly waiting for. But now we come to the part why I started writing this statement. The spring that comes to a phenomenon that makes my heart beat faster.

Artificial intelligence, in English artificial intelligence, abbreviated AI. Ai has been blamed for a winter in recent years. Nice, but not a breakthrough, and it seemed to fall into the downward curve of the hype cycle; the trough of disillusionment. Machine learning, the main principle behind AI, was found to contain no intelligence, but was only good at recognizing patterns. Chatbots and voice assistants are all disappointing and on a downward trend. Despite the promise of self-driving cars that you can order like driverless taxis, it is not even possible to get trains to run driverless in the Netherlands.


Then ChatGPT came out on November 29th and that changed everything. ChatGPT is a chatbot developed by OpenAI, funded by Elon Musk and Microsoft, among others, and its mission is to use AI ‘for good’. This does not mean that the code behind the projects is public, but that a large group of people will have access to OpenAI’s products. A statement like this is far too short to go into technical depth on ChatGPT. It is based on a model that has been “trained” on a very large data source, roughly a copy of the public Internet until sometime in 2021.

The model and the interactive chatbot software are not intelligent in the sense that the system “knows” what it is doing. There is no awareness, no memory, and no adaptation based on new input caused by users’ interaction with the chatbot. It therefore contains no learning capacity that we humans have, and therefore no intelligence of the system itself. Why am I writing this so emphatically? Because there is a deeper truth in this. A realization that forces us to think differently about what intelligence is.

The chatbot does a number of things differently than anything you’ve seen in this space before. You can really have a conversation with it on pretty much every possible level. From technology and software development, but also philosophy, politics and culture. In short: everything. What’s special is that it remembers what the conversation is about, something I personally haven’t seen before. Where it becomes special is that the output from the chatbot is original. If you put phrases into Google with double quotes, no results will come up. It therefore does not affect plagiarism, which is immediately a first breakthrough. Papers from the school will from now on be fed by ChatGPT in a way that a teacher cannot prove that the student did not write this himself. Only when a teacher starts asking questions and has a dialogue with the student can it be revealed.

A pedagogical tool

“The chatbot knows more programming languages ​​than an experienced software veteran”

But what’s more, the output of ChatGPT is simply instructive. To get straight to the point: this chatbot is based on data that may contain errors. ChatGPT comes up with insightful answers and with great conviction that they are right. Some parts of an answer appear to be completely made up with no indication that this is the case. You still need to check statements, just like with people. But don’t let this dampen your enthusiasm! The depth with which ChatGPT can think is nothing short of astounding. Even better, you can try it yourself. Within a few days, the chatbot already had more than a million users.

The chatbot in its current form has more useful applications than blockchain. Apart from writing an essay for a student, it also helps to find errors in the code. The chatbot knows more programming languages ​​than a seasoned software veteran. It is not a stretch to imagine that a good chatbot can have pleasant conversations with the elderly. Just by asking them questions about the things they say.

Model dynamic feeding?

The tool is also valuable for software developers. An actual problem I had in the Javascript frontend has been solved by ChatGPT. You can even use it to explain code or write an article on any topic, make up stories, but also write them in the style of famous authors. In many cases, you can find a deeper answer to something faster than Google.

Some call ChatGPT a direct competitor to Google. I certainly don’t want to go that far. A response from the chatbot costs much, much more than a Google search in terms of computing power. Another big difference is that the data chatGPT can rely on is old. This immediately solves one of the tool’s biggest drawbacks: its ability to process new data.

Suppose you can feed the model dynamically. For example, such a chatbot with diagnostic patient data or all the case law from recent years could improve the way doctors make diagnoses or change the entire legal industry. A system that can fight through data so intelligently and provide interpretation will truly change various industries radically.

But it’s not that far yet. It still needs to prove its practical use, and it’s not super fast. By 2023, we will see a few small iterations of the chatbot at most. Just as blockchain hasn’t made notaries obsolete, the technology behind ChatGPT also needs to prove itself, so I’m holding back. Nevertheless, I will outline the impact this will have on the software industry.

Maas and moa

“I can only have shared the terms”

With ChatGPT we get a glimpse into the future. AI is not necessarily a replacement for labor, but a tool that improves productivity. For copywriting, writing computer code, illustrations, preparation and implementation of legal proceedings, diagnosis or treatment plan. But also think about cyber security, hacking or the biological application for the development of medicines or proteins. You use people’s intuition and augment it with computing power and available data. Humans are good at intelligence, models are good at pattern recognition.

Which organizations will stand out the most? The parties that use AI in the right way and help other organizations grow faster or operate more efficiently. You take a model offered by, for example, a company like OpenAI and enrich it with other models and data. You no longer develop a tool for the support department yourself, you put it together by combining data and models. You don’t buy the models, they are extremely expensive, you pay for them as you use them. Or a model-as-a-service (mesh).

You tie these models together into a unique service, where you distinguish yourself, for example, by the data you have access to. Data indeed as the new oil and gold. The solution you ultimately offer is then based on an architecture of microservices and models. Not a service-oriented architecture (SOA), but a model-oriented architecture (MOA).

I could have already shared the terms.

A deeper meaning?

Is ChatGPT perfect? Absolutely not. But when I read about its release and saw the tool in action, I knew I was looking at something new that wasn’t there yet. If it makes one thing clear to me, it’s that the secret of dialogue and language has been cracked. This is a level that easily passes the Turing test, and that in itself should make you think. A non-intelligent system that effortlessly exhibits intelligence. Will we soon have a form of ‘artificial general intelligence’ (agi) that we don’t need to fear and that doesn’t develop consciousness?

ChatGPT works best in English, but also works fine in Dutch. And ChatGPT is the form of dialogue. For example, you have the Dall-E-2 that can convert language to graphics and photos. You don’t have to stretch your imagination what else is possible. It is only a matter of time before a system can speak all languages ​​with any voice. But we’re not there yet.


“A great future for artificial intelligence started on November 29 this year”

To stay on topic in the spirit of year-end predictions here is mine. With the arrival of ChatGPT, the cards have been reshuffled. Where AI initially seemed headed for a winter, we now see a spring. Just as the adoption of the Internet is inevitable, the adoption of AI will be inevitable. Where a voice assistant is currently still a form of micromanagement, it will evolve into a declarative interface in the coming years. You share the desired result and the assistant figures out how this can be achieved and asks you questions if it doesn’t come up. Who knows how to realize this principle will The next big thing garden in hand.

Perhaps Tesla has already entered the cards with their first robot prototype. Why? Because a robot is ultimately the basis for turning information into reality. With a good robot, you solve a lot of problems. Why a self-driving car if you can develop a good robot? Why has an AI come up with a good recipe if a robot can prepare it? Twenty percent of the Dutch population is now 65 or older, and this percentage will only increase so far. It is not rocket science to calculate how many people need to work in health care to care for this group. Without various breakthroughs in robotics, our prosperity will decline enormously. Ai is not a nice-to-have, but a necessity. A care robot without intelligence will not solve this problem.

So I foresee a great future for artificial intelligence and it really started on November 29th this year.

What do you think? Is AI with ChatGPT a breakthrough? Or a gimmick?

Leave a Comment