Let me repeat this once more: after reading this article, every bakery owner can set up a robot on their computer within an hour, a robot that will handle challenging and necessary tasks. This will not be a toy or a technical curiosity. It will be a fully-fledged office worker with extensive knowledge and efficiency.
Difficult Beginnings and Emerging Hope
Until now, implementing artificial intelligence in a production facility involved a lengthy process of analysis, data collection, selecting an appropriate AI model, and testing—in short, extremely labor-intensive work that often ended in failure, lack of results, or disappointment.
Recently, I mentioned in this magazine that the time of tedious, large-scale technological projects conducted in small private businesses, the time of implementing artificial intelligence by traditional methods and large investments, is coming to an end. I discussed how specialized programs available directly in the cloud will be downloaded to robotic devices with more or less humanoid shapes. These devices will replace humans in monotonous physical tasks.
Currently, purchasing an industrial robot involves high costs, along with a lengthy programming and testing process. It is similar to buying a car in the late 19th century. At that time, cars were very expensive, each being built from scratch by small mechanical workshops. After 20 years, with increased work automation, standardized spare parts, and technology development, cars became very affordable. The same will happen with robots. Today, industrial robots typically have the shape of a single arm performing simple tasks such as welding, cutting, assembly, or moving components in assembly halls.
Programming such robotic arms has always been a major issue. Programming was carried out in two ways. The first method involved an operator controlling the arm while the computer recorded the actions. Later, the movements were fine-tuned directly in the program code. The second method involved directly writing code, often created in collaboration with a CAD design system.
If we wanted to place a robot at reception, at the front desk, or to operate a machine, programming it would be extremely labor-intensive and time-consuming. Writing such software is practically impossible for typical data science teams or industrial engineers due to the excessive number of interfering elements. This is why, in the future, companies will emerge that will directly provide ready-made code for specific robots. If we want a baker robot, a company specializing solely in baking will lease the appropriate software. Similarly, ready-made standard software will be available for hairdressers, taxi drivers, or warehouse robots. Robots will become affordable, just like cars, washing machines, TVs, and smartphones.
The cost of robotic hardware has never been a barrier to the widespread adoption of these types of devices in industry and households. The real problem and barrier blocking the spread of robots has been their software.
Once, in the 1980s, I read a diary of an engineer building a robot: „The first month was spent designing and assembling the tracked chassis. The robot moved smoothly. The next 13 months were spent programming the robot just to make tea in the kitchen.” Mastering the mechanics of a robot is relatively easy; any engineer can make a robot that performs certain tasks, albeit clumsily. Programming the simplest tasks, such as washing dishes or reading documents, exceeds the capabilities of even the best computer scientists.
Obstacles to Technological Revolutions
Every technological revolution is triggered by the development of a breakthrough technology. Cars would never have hit the roads without the invention of the gasoline engine. The internal combustion engine enabled the development of aviation and the construction of submarines. It’s hard to imagine airplanes powered by steam engines. The invention of transistors was another breakthrough. Without the invention of integrated circuits with transistors and microelements, pocket mobile phones would never have spread. If electronic devices still operated on tube-based circuits, our smartphones would be the size of a kitchen cabinet. Similarly, the introduction of the first computers was a groundbreaking technology. Now, we are witnessing the introduction of yet another groundbreaking technology in the form of artificial intelligence, which will enable the widespread adoption of new technological solutions.
The Chat GPT 4 Assistant Revolution
Chat GPT is the first artificial intelligence model that is widely accessible. Just log in on the producer’s website (https://chat.openai.com/). The free version, Chat GPT 3.5, allows users to have dialogues with the machine. The AI model can also search for information. Unfortunately, it is not very useful because it was introduced as a kind of prototype. It is burdened with many limitations that largely prevent its industrial use.
If we decide to spend $20 per month, we can start using the higher version, Chat GPT 4. Subscription payments are made similarly to booking airline tickets by providing a credit card number. Purchasing the paid version of Chat GPT enables the creation of custom assistants.
Virtual Food Technology Assistant in a Bakery
To create a virtual food technology assistant, go to the „Explore GPTs” tab and then proceed to the „+ Create” tab. In this tab, we define our expectations for the assistant. Programming the robot is done in natural language. You simply need to state who you are and what work you expect from it. Additionally, you can define the specific stages of each interaction. Commands are written in Polish.
For the bakery technology assistant to work professionally, we should load several specialized textbooks in PDF format, along with our notes, analyses, and recipes. The assistant will have this information in memory and will use it for work.
Let’s start by defining the food technology assistant.
We need to enter the following sample text: „I am the owner of a bakery, and I bake bread myself. I need you to advise me on bread baking technology.”
To ensure the assistant uses the knowledge loaded into its memory, we should enter the following sample sentence: „Always use the notes, documents, analyses, and recipes loaded into your memory when providing responses, conducting analyses, and performing tasks related to production technology.”
Of course, this sentence can be phrased differently; however, it is essential to inform the robot that it must rely on the knowledge provided to it. If we do not specify this, the robot will use general knowledge available on the internet. Since we have indicated who we are in the first sentence, the robot will use the specific language of the bakery and confectionery industry. The assistant will also be aware of the methodology and specificity of the industry.
It is also worth adding a sentence to ensure the assistant communicates in Polish.
This concludes the programming of the food technology assistant. All changes should be saved, and the assistant should be launched. You can choose for the assistant to work exclusively with its creator. The creator can also share the assistant as a plug-in for other users. The assistant operates in all world languages; therefore, if someone in France decides to use our plug-in, „Food Technology in the Bakery,” all our notes, all knowledge, and general internet information will be conveyed in French.
Now we can ask it a sample question: „Find bread recipes that use modified yeast and contain Provencal herbs.” Our assistant will analyze 17,000 recipes across 80 books we loaded, as well as our notes. If it does not find a suitable recipe, it will suggest an alternative solution, such as recipes without Provencal herbs but with herbs from the Apennine Peninsula.