Revolution
One blogger said on his channel that “to keep up with the development of artificial intelligence is to be up to date within the last 24 hours.” The pace of competition resembles more a cycling race or Formula 1 than the rivalry of great powers, as we once had the chance to observe. Innovations appear practically every day. What reaches the consciousness of the average person is only a superficial piece of information.
For example, today I noticed that the paid version of ChatGPT introduced the possibility of conversation using a camera. It is enough to point your phone at some object, and you can have a long chat in any language about the filmed subject.
The Americans probably decided to implement this novelty after the release, on January 20, of the powerful Chinese-made language model DeepSeek, which significantly stirred up the capital markets. The stock prices of American giants fell, so they started to pull out everything they had from the closet to improve market sentiment.
A Chat with a Picture
Returning to our example of the new feature from Sam Altman’s ChatGPT. It’s very nice that we can show a language model an object and start talking about it. This is an example of a superficial piece of information. Now let’s try to understand what lies behind it.
Let’s imagine a large company that receives hundreds of thousands of cubic meters of wood. This company produces particle boards and needs an endless amount of wood that must flow into the factory in a continuous stream. Accepting such a mass of wood involves enormous risks of irregularities, fraud, and all kinds of falsifications.
And now imagine someone who, each time during the inspection of the wood, takes out their phone and turns on ChatGPT with the camera option. Previously, all documentation and knowledge gathered over many years about wood, recycling, and waste had been uploaded to this ChatGPT. Thanks to this small change, it is now possible to perform a detailed classification of the wood and assess its value.
Anyone familiar with this industry knows that this is a revolution comparable to the theory of Nicolaus Copernicus. It is a complete change in the economic system of suppliers, and it also allows for a reduction in the number of inspectors to a single person. We are now talking about a toy feature of ChatGPT, which was introduced to Europe probably mainly to calm stock market sentiment.
Revolution
I will not write about what the work of a logistician, warehouse worker, or office staff looked like in the 1990s. The appearance of the first harbingers of artificial intelligence at the beginning of the 21st century awakened entrepreneurs’ appetite for savings, improved quality, and greater process efficiency. However, those appetites quickly faded when it became clear how cumbersome, difficult, and unprofitable at a small scale it was to run artificial intelligence.
Even two years ago, the typical path to introducing artificial intelligence into a company was to collect data, hire top-class experts, perform tedious work, suffer months-long project delays, and exceed budgets. Finally, a program would appear that had to be immediately improved and refined because it did not meet the investors’ basic requirements.
In my articles, I have repeatedly shown theoretical possibilities for process optimization. These were largely theoretical considerations aimed at improving awareness of the capabilities of mathematical algorithms. However, implementing these algorithms had to involve either the private involvement of bakery or confectionery owners, who in their free time would have to explore the main issues and try to launch new technology on their own, or the hiring of very expensive—and in fact inefficient—teams and deciding on a long and therefore unprofitable investment.
And now this has changed. It has changed because of the race for leadership in the development of artificial intelligence in the world.
W-MOSZCZYNSKI ppic 3-25
Example Number 1
Undoubtedly, the form of this article may seem a bit rustic. I write this way deliberately, because I know that none of the currently operating language models write like this. Thanks to that, I avoid accusations that ChatGPT or Mistral wrote this article for me. However, before sending this text to the editorial office, I will certainly ask a model to check it.
So, what is example number one? Last month, I described how one could implement a face-detection system. I would like to emphasize that previously, the time to build and deploy such a device was measured in months. Now the time to implement this project is measured in hours—and what’s more, it is free.
Since I did it for the first time not long ago, implementation took me two days. I described all the steps in the instructions, and now this implementation would take me about an hour.
Such a quick implementation means that practically anyone can have such a device. Practically anyone today can install this system by themselves. It only requires basic digital knowledge of configuring a phone or creating websites. As we know, this is elementary knowledge nowadays, possessed by every teenager.
So what does the statement mean that anyone can install a facial recognition system? It means that any school in Poland can detect people who appear in the school for the first time. This significantly improves the institution’s security. It also means that any store can analyze customer behavior, because it can identify them. Until now, customers were identified only online, after logging into the sales system. Now this has changed.
The retail customer is identified and matched with the sales register of the cash system. Having information about one hundred thousand customers, we can create a system that will detect their preferences, needs, and even analyze phenomena and trends.
The revolution lies not in the fact that we can do this, because we could do it much earlier, but in the fact that we can do it very cheaply and very quickly. If implementing a face-detection device takes only a few hours, it will soon become widespread.
Exactly the same system can be implemented to detect and record the license plates of cars entering a facility. In the past, keeping records of entries and exits was a tedious and thankless task. Today, it is enough to install a camera and run an application that will record license plates and times. We can therefore install effective artificial intelligence within a few hours.
Example Number 2
In sales, contact with the customer is very important. In the confectionery industry, a very promising direction of development is the production and sale of cakes and special-occasion pastries. Since these are special products, they are luxurious and extravagant. A First Communion or a wedding must be unique and display the splendor and wealth of the hosts.
Selling cakes online is a challenge. Obviously, cakes cannot be viewed in person—otherwise, the whole business would be unprofitable, as cakes have a very short shelf life. So, we rely on photos. Of course, with orders come the proverbial one hundred questions about dimensions, harmful ingredients, and other matters that in the long term can irritate anyone running such a business.
It would be ideal if someone were delegated to answer questions such as, “Do the candles contain potassium chloride, because it’s said to be carcinogenic, and after all, they smoke when blown out?”
To provide quick answers during flight bookings on web services and in retail sales, automatic chats appeared—the so-called bots. However, bots by nature are simple and respond in a basic, often childish way. When a question is too complex, they say they did not hear it, or send a generic comment like, “Yes, I completely agree with you.”
A few years ago, creating a decent bot whose intelligence equaled that of an average five-year-old took several months and required a large group of specialists. This field was called Natural Language Processing (NLP), which included specialized libraries detecting syntax, letters, sentences, and various other important elements of spoken communication.
As we know today, we can talk with various types of chats—GPT, Mistral, DeepSeek, or Gemini—in the same way we talk with a living person. Moreover, these chats are free, and their knowledge, intelligence, number of languages, and ability to understand unclear spoken words are astonishing.
So, wouldn’t it be nice to have such a “smarty” on the cake store’s website? Then, a question about the harmfulness of sodium citrate potentially present in some ingredient would not cause such consternation.
Language models simply love answering such complex, sometimes tricky or error-laden questions, going into details and even showing examples or chemical formulas. I’m not sure whether the word “love” is appropriate for these devices, but they are undoubtedly specialized for it.
And again, we have a revolution. Within a few hours, we can implement such a chat directly on a website. It does not require programming skills—just technical knowledge of configuring a phone or a router. Practically every active person has such knowledge.
As with devices reading license plates or detecting faces, it is enough to log into a free cloud account, for example AWS or Azure, play around a bit, run the proper language model, and connect the local or cloud-based computer that handles the cake shop via API protocol. It’s a task comparable to connecting your phone to home Wi-Fi.
Example Number 3
We are becoming increasingly dependent on external companies that provide us with cloud services. In the previous example, I showed how easily we can introduce artificial intelligence into our lives—an intelligence that exists somewhere in an analytical center with enormous resources and data processing power. A center that, at some point, can cut us off from its services and make us return to the efficiency of the 1960s.
It must be admitted that the digital revolution is much more flexible than previous ones. I mean the revolution of electricity, the steam engine, and the computer revolution. In all these cases, cutting off technology meant a quick and inevitable return to earlier times.
Let’s imagine such a situation not today, but in the near future: our confectionery is equipped with various tools that closely use artificial intelligence. We analyze employee working time based on face identification at the gate. We analyze the number of positive and negative comments under our videos praising our confectionery products. We participate in discussions and handle customer service entirely without human involvement.
By the way, I should mention that this week Sam Altman’s company introduced the so-called “Operator.” This device, in simple terms, is an invisible robot that sits in front of our computer and handles all our matters, such as bank transfers, correspondence, order completion, and sending them to the production process. It’s something that completely replaces an office worker. It can do accounting and order pizza, settle transactions, and chat with our friends on our behalf.
So, returning to our hypothetical automated bakery-confectionery equipped with the latest technologies: what will happen if the providers of all these cloud services cut us off? The answer is—it depends on us.
Model Code Is Free
When we talk about the revolution of electricity, if the power plant cuts us off from electricity, we return to the dark ages. It was similar with the mobile phone revolution. This revolution, however, has the advantage that most solutions are based on open-source code.
This means that anyone can download all or part of a language model. Downloading an entire model for personal use is not very efficient, because such a model must have extensive infrastructure.
Returning to our example of the confectionery of the future: if someone wants to secure themselves against various contingencies, they can simply download a language model onto their own computer. There are many language models that can be directly downloaded and used without cloud services. Practically all language models available online can be downloaded in compact form and installed on your own local device.
If we want our own language model to handle the cake shop, we can limit ourselves to a model that only handles conversation. It won’t recognize images or code, but such a model will work flawlessly on a local device. Moreover, we can train and specialize this model in our products. We can feed it recipes, photos, dimensions, and all the information that could potentially appear in customer inquiries.
The same applies to systems recognizing silhouettes, faces, cars, or analyzing goods movement. All these solutions can be downloaded for free to local devices, used without limitations, and bring unlimited benefits. If this isn’t a revolution, then what is?
Summary
The digital revolution brings significant changes to our lives. For now, we can still sit on our couch and ask free Chat Gemini, for example, how to fertilize fruit trees in the garden. Meanwhile, sooner or later, tools, devices, and systems will appear that will significantly improve the economic efficiency of individual businesses.
This rule will also apply to our industry. If someone introduces innovations that increase competitiveness, we will quickly feel it in our wallets. These changes will be introduced personally by entrepreneurs—or possibly by their teenage children. The range of emerging innovations is impressive.