Future technologies in games: they really needed?

Computer games have been entertaining and educating people of all ages and generations for many years. Different games do this in different ways. However, there has been a certain stagnation in the player’s interaction with the 3D world. Moreover, people are trying to convince us that neural networks are suitable not only for generating images or texts. Today we will see what developers are preparing for us in the future and immediately vote on whether we, ordinary players, need it.

New types of interaction with NPCs

TiGames showed a small demo using voice input to talk to an NPC. The player approaches the character, starts talking to it, and it responds and does what they want. For example, they can ask a merchant to show their goods or a hairdresser to cut their hair in a certain hairstyle.

How does it work? First, a human voice is recorded and immediately converted into text. After that, the linguistic model breaks down the words, and based on this information, AI determines what exactly the NPC wants from it. After that, it needs to select the required query and generate an answer (in text and voice formats) to the question. Of course, the NPC’s face animation is also generated. Therefore, some of the AI choices should already be pre-programmed.

A good joke about the interpretation of the acronym AI — Animal Intelligence.

The video also shows an example of classic photo generation. However, this is not just an image, but a new texture on the main character’s spaceship. This opens up an interesting idea of customizing the same cars, for example, in future installments of Forza or NFS.Such interactivity should be well suited to RPG games for playing various roles by the user. For example, coughing into a microphone to ask for help from an NPC doctor. AI will not react like an ordinary customer but should offer emergency assistance. Or at least provide a discount on services. Another example: during an argument, one of the teammates should raise his voice sharply. Then AI could accept or reject the player’s arguments.

In the game Mecha BREAKshowed the ability to communicate with one of the characters for a short briefing on the upcoming mission, select a humanoid robot and change its color scheme. All this was done by voice.

To increase interactivity, you need to teach NPCs to tell you more about the reasons for quests. That is, it will be able to diversify the tasks «Go and bring 5 flowers» in «On a mountain hill, on top of a rock guarded by a goblin pack, you will find a flower. It will help heal my leg from the burns». Of course, you need to keep the AI within the framework of the general world story, without adding unnecessary information.

Can you imagine what will happen if a gamer says phrases in real time «Surrender!», «Hands up!» etc.? This will significantly increase the level of immersion in the game. This will be especially noticeable and useful for streamers and their viewers. Of course, it will be necessary to make sure that players do not shout into the microphone for the whole house. To do this, you need to ask to increase the maximum sensitivity of the microphone for a calm speech, without shouting and emotions.

It is clear that there are three main problems with this symbiosis of voice, text, reaction, and animation technologies:

• Security of personal data. More precisely, your voice. Therefore, it is necessary to have offline versions of voice, linguistic models, and AI itself
• Additional performance costs for data processing are required;
• To combine the previous points, you need an expensive PC. Consoles are out of the question for now.

Most gamers are already used to coming to the same merchant, pressing F to open the dialog, and selecting Trade from the menu. And here, developers need to solve a simple question: is this a necessary thing that is worth spending their time and money on? The question is quite controversial, so a special vote awaits you at the end of the article.

«Honey, AI is killing our children»— Improvements to Bots

Gamers have long wanted not only good graphics but also bots. Bots are used in many computer games, but their main role is as virtual opponents for real players, imitating strong, smart, and experienced warriors. However, the main problem with such «enemies» — is programming their behavior for different game situations. If the complexity can be programmed somehow (for example, the speed of reaction to a player getting into the bot’s sights, pressing the trigger, or WallHack), then the variety of situations — how much energy, enthusiasm, and time the developer will have.

One of the best examples of the implementation of bots for first-person shooters is First Encounter Assault Recon, or simply F.E.A.R. (2005). Bots are not only deadly, but also terrifying because of their constant radio conversations and attempts to surround the gamer. If the player cannot be surrounded, they will first try to throw a grenade and then ambush the player.

AI is helping to solve this really difficult task. In this case, it is a raid boss in MIR5. On the one hand, it will be an interesting experiment. On the other hand, it has an unknown outcome. There are fundamental problems in the use of AI in raids that developers must solve.

The first problem is the pattern (template) of boss attacks. If they are set only by the developer, all boss actions will be highly predictable. But the first attempts of ordinary gamers to defeat the boss, even after seeing his attacks on Youtube or reading guides, will be unsuccessful. First, a person needs to get closer to the boss: learn the timings, determine the best moments to block, deflect an attack, or use an iframe (invincible frame). In addition, you also need to attack and execute the boss mechanics yourself. Of course, all this training will take some time.

Now, let’s take AI that has gone through dozens or hundreds of battles with developers/testers against ordinary players who have only seen it in guides. The result is inevitable — humans will lose.

Let’s return to the pattern itself. Raid developers must think of, write, program, and animate all the attacks of the raid boss in advance. Let’s say there are six attacks: front left paw, both paws at the same time, tail, fire from the mouth, jump on someone with the whole body, unblockable AoE (Area of Effect) on half the map with a state. In normal raids, all of these attacks would come one after the other. A raid party of average gamers will learn this pattern in about 30 minutes or an hour and come up with a counter to AoE by casting a massive iframe on all party members.

What will happen if AI «is set free by» and only demands to win? It will start spamming AoE until the entire raid dies. In the video, we see an example of a similar situation: the boss jumps on a distant player and can attack the entire group with one attack — oneshot of the entire party. The first players will know the exact list of attacks, but those who meet the boss will then see an ultimatum strategy to destroy the party. If people find a way to counteract this tactic, AI will come up with a new one.

The battle between machines and humanity will begin! Would you like to see Malenia from Elden Ring attacking Waterfowl Dance non-stop? That’s right, without a window to attack. Why? The AI is alive as long as it has HP, so it can’t lose it either.

There is one interesting point. The class system in RPGs has a special class for clearer control of the boss. That’s right, the Tank, which has the Aggression function to «tank» all enemies. Usually, this class is taken by an experienced gamer with a lot of experience in such games. It is not known how AI will react to the Aggression skill. However, the trailer showed that the AI jumped on a distant player on purpose. And this already breaks the concept of an entire class in MIR5

After reading the arguments, you might think that the use of AI in bot creation is — bad. Of course, it is not. That is why we propose a vote:

The application of new technologies led by AI in games can lead to a gameplay revolution. However, do players want it? Neither developers nor gamers know for sure. We will not know the results of such an experiment until we try it. We hope that the technologies shown will not only look good in videos but will also be used in games