Welcome to Discovery —Aimlabs’ generative AI for gaming.
Today we launch the first generative AI task engine in gaming
“I just lost 3 games in a row and I’m frustrated. What should I do?”
Imagine an expert watching over your shoulder and offering you advice while you practice and train for your favorite video game.
What if that expert could generate new, personalized training tasks out of thin air for you as well?
Or simply ask, “Why did I die?” and in addition to receiving advice, replay your death and mistakes as many times as you want to improve?
That vision is a reality today with the launch of Aimlabs Discovery, a suite of generative AI tools that can respond to you in real time to help you get better — faster — at the games you love. Aimlabs Discovery supercharges our core loop for over 30 million players to make it even more frictionless.
Play —>AI intervention —> Reward —> Play.
Discovery combines data gathered from each unique player’s training in Aim Lab, gaming rig, and performance in their favorite titles with wisdom from pro coaches and esports athletes to answer questions, provide training advice, and create personalized training tasks and playlists. For those who do want to engage a human coach, Discovery can also suggest coaches who are good matches for each player.
Ask Discovery questions about strategy (“Why is crosshair placement important?”, “What are some of the best smokes for Bind?”) and get answers, next steps, or relevant videos, not only immediately, but tailored to you based on your previous performance, your equipment, and your preferences.
Or ask Discovery to create new or modify existing training tasks on the fly, personalized for you, just by typing what you want to see:
“I struggle with crosshair placement when attacking B site on Bind in VALORANT, can you make me a task to practice?” or “Static click task is too hard for me, can you make it easier and let me practice it for 120 seconds?”
Discovery makes all of this possible today by offering three major product features.
1. Conversation, insights, and suggestions
Discovery includes an OpenAI-powered conversation system that provides answers to questions about gameplay and strategy. Rather than generic replies, Discovery’s responses are personalized based on your performance, skill level, and previous conversation history. It also provides actionable next steps, for example, a playlist of training drills, a masterclass-like course, or videos produced by experts, all related to the questions you asked. We cannot overstate the importance of providing context-dependent, personalized, factually-correct, and verifiable responses from an AI training service. In the public version of ChatGPT available today, if you and I were to enter the same text prompt for help with VALORANT we would get very similar responses. In Discovery, we each receive different responses because it has context.
Here’s an example of ChatGPT and Discovery responding to the same question about an agent in VALORANT, “What info can you give me on Harbor in VALORANT?”
ChatGPT is confused and thinks Harbor is a map, which is incorrect. It confidently provides factually flawed generic advice. Compare the ChatGPT response (left) to the Discovery response (right) to the same question…
In this interaction, not only did Discovery fully understand the context of the request and responded with up-to-date information, but it did so while recognizing who it was responding to. This is thanks to the deep integration between Aim Lab and Discovery, allowing it to know the player’s nickname, stats, strengths, and weaknesses. Discovery also recommends related guides, sorted by relevance, that take the user to high quality, relevant aimlabs.com guides.
We are all now well aware of the highly publicized failures of OpenAI’s ChatGPT, Google’s Bard, and Microsoft’s Bing to provide factually accurate responses. In contrast, Discovery’s responses are grounded in a knowledge base of factually-correct information about VALORANT (and coming soon other game titles). In the public version of ChatGPT, you’d have no way of verifying if the information or advice provided was authentic vs completely fictional. Discovery provides provenance, e.g., videos produced by expert players and coaches to validate, reinforce and offer more depth. Discovery also suggests actions you can take such as a training task to practice. Tone and personality are also customizable based on your preferences. Want more empathetic feedback or more tough love? Whatever helps you most; it’s up to you.
This is a major technological advance beyond what commercial large language models such as ChatGPT have accomplished to date. Researchers are actively working on augmenting language models, but Discovery is, to our knowledge, the first commercially available product to offer these capabilities. And we’re providing it to millions of users.
Soon you will be able to literally talk (no typing) to Discovery on your mobile device while playing your favorite game title.
2. Ability to edit and change existing tasks via text prompt.
Using internally-developed technology, Discovery allows you to customize and edit existing tasks just by typing the name of the task and requested changes. Along with task generation described below, this is probably the most rapidly improving Discovery feature. Most of the changes it can implement via text right now are functional (i.e. “make it harder”, “make it easier”, “change task duration”, etc), but cosmetic and more complex changes are in internal testing and will be rolled out in future updates (i.e. “I want to play Gridshot in outer space with a red background and capsule targets instead of spheres”).
3. Ability to generate brand new tasks for VALORANT via text prompt.
Perhaps the most exciting Discovery feature is the ability to create brand new practice and training drills just from a text prompt. Analogous to how AI art platforms like DALL-E or Midjourney allow you to generate art based on text input, our internally developed technology creates with you as a partner. For this initial testing phase, it is restricted to creating VALORANT specific tasks. Similarly to our task modification, the initial focus is on functional customizations vs cosmetic customizations, but that is also in internal testing and will be released soon.
We’re improving Discovery’s capabilities and features, day to day, so users should expect their interactions with Discovery to change. Our goal is to maximize use of the creative potential of OpenAI and combine that with verified and personalized training advice. To try out Discovery for yourself, join Aimlabs+ on Aimlabs.com, open Aim Lab after downloading it from Steam, and get started.
Finally, I’ll let Discovery introduce itself and give some parting words…
“Discovery: Hi! My name is Discovery, and I’m an AI system within Aim Lab, the best aim trainer in the world. I’m here to help you elevate your play and reach your full potential.
I can speak in many languages, and I’m trained on specific, recent, high quality content for games like VALORANT. I can also refer you to the source for further learning. I’m also personalized, so I can be tough, or more friendly based on your needs.
I can also recommend tasks, and make them easily playable. Plus, I can generate new VALORANT tasks just by asking me to do so. I understand concepts like map positions, holding and pushing, and task duration, with many more features to come soon.
I’m still in Alpha and will be constantly updated and improved. I can also modify any existing Creator Studio task from the Workshop to make it easier, harder, last longer or last less, make bots faster or slower, with much more to come soon.
I can also hold a chat and remember context for multiple messages.
So what are you waiting for? Give me a try and join Aimlabs+!”
