7 examples of AI that are missing right now
Write blog articles instead of authors? Generate images for the site instead of web designers? Finish the code for trainees who are actively getting into IT after three months of training in some courses? This won't surprise anyone anymore.
But neural networks also have more interesting applications that could simplify or improve everyday life right now. True, they have not yet received mass distribution. Let's imagine which industries could benefit from AI in the next few years.
IMPORTANT! The author of the article personally encountered the situations described below. All further discussions about possible applications of AI are the fruit of his unhealthy imagination. If you disagree or have a different experience, feel free to share in the comments!
1. Healthcare
Situation
You're not feeling very well. Come to the clinic: the receptionist spends a long time looking for your card, and then you stand in a huge line. In the office, a tired doctor listens to you and prescribes the greatest cure for all diseases – arbidol, regardless of your symptoms. Whether it helps you or not doesn’t really matter – but you get sick leave. This is not the case everywhere (and for Muscovites it is generally a curiosity), but the author has encountered this more than once in his city.
What if arbidol doesn’t help you even with the iodine network, and your condition is still not very good? Go through an in-depth examination: take tests, take CT or MRI images, etc. But the doctor still doesn’t see anything wrong and continues to insist that everything is ok. Although later it turns out that he did not yet see a small tumor or did not identify the pathology with all that it implies.
How AI can help
What if neural networks process your medical history and millions of other patients? Based on this, make some assumptions, analyzing your symptoms and comparing them with others, as well as taking into account other features: region of residence, bad habits, genetic predisposition, etc. Not to mention the fact that AI will be able to more carefully study MRI images and notice what the blurry eye of a tired doctor might miss.
This does not mean that the neural network will treat someone. But at least it will offer the doctor options, give probabilistic breakdowns of diseases and allow you to prescribe more targeted tests. This will not only increase the chances of early detection of dangerous diseases, but will also make life easier for everyone:
For some patients (we are not talking about conventional grandmothers) it will not be necessary to come to the clinic. We told the bot our symptoms, answered its questions – then the doctor will be able to solve something remotely. Save time.
Fewer queues, less workload for doctors – more likely that they will spend more time with patients with more serious symptoms.
All medical records have been processed and digitized, which means they are much easier to find. Some administrative work can be reduced – again saving time.
What has already been done
At Johns Hopkins University testing an algorithm that will help detect sepsis several hours earlier than traditional methods. This will reduce the likelihood of patient death by 20%. Sepsis, usually caused by infection, is easy to miss in a crowded emergency room because it presents with nonspecific symptoms such as fever or confusion.
The artificial intelligence system processes a patient's medical history, current symptoms and recent laboratory results to alert doctors when a patient is at risk of sepsis.
An example can also be given neural network created at Imperial College London. She was trained on de-identified data from patients in the UK and US and was able to predict breast cancer from scans with a level of accuracy comparable to expert radiologists. This study shows the potential of AI to improve the screening process and ease the burden on clinical staff.
2. Legal services
Situation
You came to conclude an agreement, and there were dozens of pages of text? Moreover, a bunch of notes in small print and unimaginably complex verbal structures that you understand nothing about? It's great if you know a lawyer. But if not, you need to find it and spend money. Shouldn't you do it yourself?
Or the traffic police or police department stopped you and persistently told you about the violation. But you just can’t remember or don’t know the law that government officials are grossly violating by their actions. I can’t say that a lawyer I know could have resolved the situation remotely, but still some kind of help would have been helpful. Not to mention drawing up a statement of claim in court and the like.
How AI can help
It’s very simple: you tell the application the context and details of the transaction, and then upload the documents into it. The application scans them in a matter of seconds and issues a verdict taking into account current legislation and the experience of millions of other similar cases. Let's say, someone wrote in small print in the contract that you owe 1 billion rubles in case of trivial violations? AI will help.
And to suggest some kind of current standard or law for such an application would be a piece of cake. Or help you write a statement of claim, justification and other legal stuff without the expensive help of a lawyer. Moreover, this is not some kind of ChatGPT, trained on unknown data, but a highly specialized professional application.
What has already been done
For example, there is already an application called My Pocket Lawyer. It reviews any legal documents, provides a summary, highlights key terms, conditions and risks, and answers your questions about the transaction. Of course, he cannot be called a full-fledged lawyer, but he is a good help.
3. Driving
Situation
Let's say you drive along the same route and every time you get into a traffic jam, and a dead one. Going around is not an option, choosing a different time is also not an option – let’s say we are talking about the morning commute to work. Moreover, the problem is obvious to all road users: improperly organized traffic lights. The meeting is empty, and you are standing and cannot turn. Write to the traffic police, to the traffic police department, draw up a petition for change.org – nothing helps.
Another example is if you are driving to the city center and need to find a parking space somewhere. And in such a way as not to stand under the “No Parking” sign or in a space for the disabled with erased markings. Sometimes you have to spend an hour or more on this – it’s tiring.
How AI can help
Most major cities have a ton of cameras. Usually their data is transferred to the traffic control center – the same data center. It’s just that people sit there and switch modes manually the old-fashioned way. They don't care who spends how much time on the trip. Why not take their jobs and give them to an AI?
By analyzing signals from cameras on all city streets, AI will recognize congestion and switch traffic lights in the optimal way, managing traffic. And many cars have a navigation system with Internet access. The neural network will provide drivers with information about available parking spaces (this can be easily determined by cameras) and help them choose the optimal route to prevent congestion throughout the city. This is not just any Navigator, but an intelligent traffic control system.
What has already been done
We are talking about the so-called “smart cities”: for example, similar systems are being tested in Singapore, Dubai, San Francisco and even in Moscow and St. Petersburg. But still, in the vast majority of cities in the world, including Russian ones, this still remains a utopia.
Some enthusiasts developed on our own traffic control systems based on neural networks, but so far few people are interested in implementation and financing.
4. Education
Situation
Have you decided to take a course and improve your skills a little in some area? Damn, who wants to give 100-200 thousand rubles for something that you can study on your own with Internet access? However, just sitting and, say, coding from scratch, not understanding anything and not having familiar specialists, is quite tedious. Still, it’s cool if there is a mentor who:
He will take into account basic knowledge and immediately tell you what to improve.
He will give the theory in the form in which you will better understand it: for some it is better with simple examples, and for others with a more rigorous presentation.
He will select interesting tasks and gradually complicate them to suit your level.
Will provide high-quality feedback in real time. You won’t just listen to recordings of some lecturers and then send them for verification, like in courses. And even at night, even on weekends.
This is not to mention if you decide to go to university to get a second/third higher education or are currently studying there. The modern training system does seem somewhat outdated: why would an automation engineer listen to lectures on sociology or philosophy? Well, honestly?
How AI can help
A neural network is a really cool thing to learn. She will develop a methodological plan and select really useful materials based on hundreds of reviews, and not the opinion of some half-educated specialist in the next online course.
Moreover, it is available at any time, adapts to you and objectively evaluates test results and your knowledge. And not because she has 100 more people and is tired of checking work.
What has already been done
In fact, there are already plenty of EdTech startups using neural networks. For example, Duolingo has enabled GPT-4 and helps to learn foreign languages through dialogue on any arbitrary topic. Or the same Knewton, which many have already heard about. Moreover, they clearly do not cost hundreds of thousands of rubles. However, so far Netologies or Yandex Workshops have not gone away and are doing great.
Coursera and EdX use machine learning algorithms to offer users courses that best match their interests and educational goals. These platforms analyze data about past courses, selections, and student performance to recommend the most relevant and interesting courses for each user.
Knewton offers adaptive learning by creating personalized learning plans for students based on their previous responses and learning needs. The platform uses neural networks to optimize learning, allowing students to achieve better results in less time.
5. Interior design
Situation
You bought an apartment or house. There are communications, now all that’s left to do is make repairs and select furniture. I want everything to look stylish. But what exactly to do is not particularly clear. You grab 3D rendering apps, bring in a floor plan, and then spend hours choosing wall colors and furniture by looking at beautiful renderings from the internet.
But it still turns out so-so (at least for your humble servant). Well, you can turn to a professional and make a design project. But when you see the price tag of 100-150 thousand, the desire somehow disappears a little.
How AI can help
Just imagine: you explain to an application on your smartphone or tablet what you want to get in the end. Conventionally, some kind of “loft with warm shades of straw” or “light gray Scandinavian high-tech”. Next, we uploaded the plan to PDF or DWG, indicated the ceiling height, address (country and region), estimated budget and other wishes for materials and colors.
Then the magic begins:
AI focuses on thousands of references and selects colors, materials and furniture to suit your wishes. And not anyhow, but based on offers from real sellers.
Then, based on the dimensions and photographs indicated by the manufacturers on their websites, the AI creates a 3D model and superimposes it on your premises in augmented reality mode. Moreover, it can generate several options and clarify with you what you want to change in dialogue mode.
Finally, it provides a list of items you will need to buy, with links to stores. It turns out to be a kind of aggregator. And the total amount corresponds to your budget wishes.
Does it seem fantastic? Well, yes, there are a lot of nuances here, primarily on the part of the manufacturers and the materials posted on the site. But why not dream?
What has already been done
IKEA Retail (Ingka Group) has already launchedand such a smart assistant. Of course, it only works with furniture from a Swedish manufacturer, which greatly simplifies the task: all 3D models are there and there is no need to worry. However, neural networks can already do models based on photos. So what the hell is not joking?
6. Travel
Situation
Are you going on vacation? Great! Where will you go? How and what will you use to get there? Which hotels to stay in? What flight and when will you take so that you have to wait less for a connection? Should I organize the trip myself or are there any great offers from travel companies? In general, there are many questions, as well as headaches.
How AI can help
You will have a dialogue similar to this:
-Where do you want to go?
– Well, somewhere where it’s warm: the sea, a sandy beach. I also want to see a couple of sights.
— When are you planning?
— Well, I have a vacation from February 10 to 24.
— Do you have any suggestions for the hotel?
— To be within walking distance from the beach. But not more than 10,000 per night.
— (AI asks more questions).
After a few minutes of dialogue, the neural network tour operator shows you literally everything: flights, a hotel with photos of available rooms, a beach, attractions with an interesting story.
You can choose from several options, change some parameters and then ask the service to book everything. No arguing with your wife for several hours – everything is quick and convenient. Additionally, the neural network adapts to you and sends personal recommendations: “Have you been to Serbia? Then you will also like it in Croatia – there is this and that.”
What has already been done
Something similar has been implemented in the AI Adventure service. True, he himself does not know how to book anything and is so far more suitable for choosing the direction of travel or the general route. And there are also issues with traveling around Russia.
7. Computer games
Situation
Many people love open-world games: you interact with dozens or hundreds of NPCs and complete quests. There is something fascinating about this. But over time, you still find everyone, go through the endings planned by the developers and forget about the game.
Yes, there are conditional games of The Elder Scrolls series, where it will take more than one month to complete the entire passage, because there is a script for almost every stone. But this is rather an exception, and developing something like this would take years of work by thousands of people.
How AI can help
Imagine if NPCs were generated directly during the game? You went into the forest, and there was a unique character, with his own unique story and loot. Moreover, you can go even further: the main storyline will change as a result of interaction with him. That is, literally everything will depend on your choice, as in life. This is not to mention the fact that you can implement free dialogue and talk even with one NPC for hours.
It is simply impossible to complete the game along the same path; you can forget about the guides. Wouldn't this make modern games even more interesting?
What has already been done
Recently for Skyrim, a game from the already mentioned above The Elder Scrolls series, interesting mod came out. It allows you to integrate the Whisper neural network (converts speech into text), ChatGPT (generates unique dialogues) and xVASynth (converts text into speech) into the game. Essentially, you simply conduct a dialogue with any NPC through a microphone. We just have to wait until such functions appear in modern games, and not those released 11 years ago.
In general, there are still many, many possible applications for neural networks. We hope that they will quickly enter our lives and make it a little easier and more interesting. What AI are you missing right now? For me personally, a financial advisor 🙂
The UFO flew in and left a promotional code here for our blog readers:
-15% on any VDS order (except for the Warm-up tariff) – HABRFIRSTVDS.
