Powered by

# What is AIGC

AIGCs represent sophisticated, interactive artificial intelligence athletes characters, specifically created for sports training. The function of virtual coaches is to aggregate several enabling technologies in order to propose an integrated and personalized approach towards sport activities

# ViT based computer vision

  • Real-time analytics of the athlete's movements and technique
  • Recognizing correct exercise performance and identifying potential errors.

# Decision Intelligence

  • A flexible system, where possible, adjusts the training plan according to current performance and progress.
  • Ensuring optimally intense exercises and types for the best training outcome.

# Reasoning agents

  • Coaching feedback using a team of domain-specific language models pre-trained on sports and training techniques.
  • Contextually relevant guidance and instructions during training.

# Real time communication

  • Human-like voice interaction: the athlete not only receives instructions but can give feedback during the training.
  • Speech recognition: Communicating with athletes in their native language.

# Sensor and device integration

  • Aggregating data from wearable devices and sports equipment for more detailed analytics and performance.

# Gamification:

  • Integrating game elements in the training process to enhance motivation and interaction

# Personalization

  • Considering the specific needs of the individual athletes, their distinctive characteristics, goals, and preferences in designing and refining the training programs.

AIGCs operate as an integrated system where all components work synergistically to provide athletes with a unique training experience that combines expert knowledge, technological innovation and personalized attention.These virtual trainers are capable of not only providing instructions, but also motivating, adjusting and adapting to the needs of each user, thereby creating a new standard in the field of sports training.

Meet ACTIcat — an intelligent interactive coach based on Actiquest Sport Knowledge LLM.

# What does the ACTICat do?

  • Analyses the athlete's movement in real-time;
  • Develops your training plan and makes adjustments depending on the situation;
  • Provides personal recommendations for the correction of exercise technique.
  • Flexibly adjusts the volume and types of exercises depending on your health and mood;
  • Provides recommendations for nutrition and recovery, considering the peculiarities of your body;
  • Motivates with $ACTI tokens for accomplishing training goals and personal records upon completion of the workout;
  • Speaks with a voice similar to A. Banderas or A. Mironov, depending on the language;
  • Empathic & emotional communication with athletes;
  • Follow your mood;
  • He himself may be out of sorts-sometimes very rare, imitation of fatigue.
  • Conducts a system of quests and missions for more enjoyable training.
  • Joins physical training with mental practices for the all-around development of the athlete, i.e., able to switch to psychologist mode.
  • Consider long-term objectives, environment and weather, solar activity, your biorhythm, and the state of sleep on the previous day, and incorporate the short-term objectives when developing your training program.
  • Analyzes health and physical condition data with a view to enhance your workout session for better performance and injury prevention.

Rather than being coaches who give instructions, AIGC in Actiq become veritable companions in the sports journey of action undertaken by each user. Advanced technology, a personalized approach, a reward system, and friendly interface combine in an environment where training becomes entertaining and motivating. This is a new standard in sport preparation where technology serves to unlock the full potential of every single athlete and make the path to achievement effective and truly exciting.

_Opposed to humanoid robots from Will Smith starring Sci-Fi thriller “I Robot”, cat-like AIGC character design could make the whole process of learning and training more fun-less intimidating. Optimus robots, Cybercab robo-taxis, and Robovan buses were completely inspired by this film, which means that the marketing of robotic innovations is surprisingly predictable.

Why a cat? Humanoid robots sometimes fall into an "uncanny valley" where they are almost human but with subtle differences that are disquieting. The virtual cat character avoids this problem in its entirety by taking on a very non-human form that is universally attractive and non-threatening. People have a natural affection for animals, cats especially. This helps, in some cases, to work out a sense of camaraderie and, therefore, motivation to make the workout experience more comfortable.

It is possible to attribute a personality to ACTIcat, adapting it to different age ranges, sports, and forms of training in such a way that the AI will adapt to different users' needs.

Thus, it might be that athletes get emotionally tied with the sweet and friendly AIGC; hence, commitment to the training program increases. Therefore, with this, it will keep them more motivated by sticking to your exercise schedule.

Unlike aggregators that offered remote classes with human fitness trainers, such as Fittr-which came during Covid times, or applications leveraging machine learning in endurance training apps like PKRS.AI and Enduco, Actiq's multimodal AI changes the game in sports training, taking on several at once.

AIGC always stays in touch with you. Using the Grade Adapted Pace model with a strong feature of speech recognition, the AIGCs develop extremely adaptable, personalized training plans for sports enthusiasts according to their particular goals, fitness level, and progress. This would better reflect the effort of the athlete and guide their training effectively through the use of game-based training blocks. AIGC continuously learns from the athletes through the processing of health data originating from a range of sources, such as Strava, Google Fit, Garmin Health, among others.

# AIGC in 8 bullets (to avoid a long read):

  • User-configurable virtual characters;
  • Predicts and monitors training outcomes;
  • Analyzes physical/mental state;
  • Provides motivated, personalized feedback within training blocks.
  • Sees the trainee using 60fps ViT and estimates the quality visually .
  • Constantly readjusts the training plan by set goals;
  • Tracks endurance, pace, threshold, V02Max, and anaerobic parameters;
  • Manages monetary rewards upon completion of training blocks and quests.

# Can AIGCs coach better than humans?

AIGC will soon surpass human sport coaches, and here's why:

Data Analysis: Human trainers are unable to analyze large amounts of data as fast as AI would. Probably one of the main advantages of AI is high speed in processing and analyzing big volumes of complex data about the health status of athletes to make better decisions and improve the process of training.

Availability. Unlike human trainers, who are restricted to their personal time and schedules, AIGCs don't need breaks or days off. At any moment, whichever the time zone or other restrictions, they are available, providing necessary support or feedback continuously. AI trainers are always in touch to help.

Subsequence. The human factor is reflected in the stability of the training process: human coaches may be in different moods, have biases, or other subjective factors. AI overcomes these shortcomings by offering a consistent coaching approach based on rigorous algorithms. This means that the quality of your training will always be consistent.

Long-term tracking: AI has surpassed the capabilities of human coaches in tracking athlete performance over time. It still proves hard for a human even with the use of modern applications in data management. AI is constantly analyzing and adapting the training procedure with long-term data that shifts the athlete better toward his goals.

Scalability. AI coaches can work with as many athletes at one time as an individual wishes, which is impossible for human coaches working one-on-one or with groups. That is why AI has become an indistinguishable tool in mass coaching.

Flexibility. AI can make the training programs personalized to the needs of each athlete. The immediacy with which AI adapts in real time, based on the changing new data inputs, is second to none. AI by AI Program in a part of decentralized AI model training will continuously enhance AIGC’s into state-of-the-art, adaptive, and effective in any environment.

Neutrality. AICGs have no personal biases and psychological factors that can influence human trainers' decisions. That is to say, there is an objective approach towards training: every athlete is treated on an equal footing, being effectively motivated to achieve whatever they aim to.

Availability. AI sport coaching is usually more accessible and cheaper as compared to classes with human coaches. This offers a chance of professional coaching to more beginner athletes.

Economic accessibility. That is an economical benefit for AI as well. For example, an hour's training with a tennis coach costs on average US$100, while services of the AI coach are 100 times cheaper. A subscription for AI coaches in different sports will cost about $20 per month, which will allow everyone to develop healthy habits and achieve high results.

AIGC's are always on duty. By implementing GAP (Grade Adjusted Pace) models and robust speech recognition, AIGC's can create highly adaptive, personal training plans for non-pro athletes, keeping their unique goals, fitness levels, and progress in mind. It gives a more accurate reflection of an athlete's efforts and guides their training effectively via gamified training blocks. AIGC's are constantly learning from athletes using realistic NLP engine and by processing athlete's health data coming from various sources such as Strava, Google Fit, Garmin Health, etc.

# AIGC in a nutshell (backend side)

# Vision Transformers

Sports-related CV models running on mobile devices, CDN Distributed, open source, decentralized. Applied technologies: modified NAS Yolo Pose. Now migrating to "Sapiens" ViT)

Vision Transformers (ViT) allow us to analyze users' movements during training with high accuracy and a speed of 60fps. ViT models work on the principle of transformers and are fully integrated into language models. They divide the image into small areas (patches) and process them using attention mechanisms. This makes it possible to more effectively recognize complex movements and postures of athletes;

  • ViT can more efficiently process raw video on user devices than convolutional neural networks (CNNs) that are typically used in computer vision systems.
  • ViT can recognize complex sequences of athlete movements. This is especially useful for improving exercise technique in real time, offering users more accurate and personalized recommendations from an AI coach.
  • ViT can recognize the smallest details in images, allowing for better recognition of fine motor skills and recording deviations in exercise technique. For example, the slightest deviations in tennis or gymnastics can negatively affect the final results (in competitions).
  • ViT requires less resources for training and video processing (output). The video no longer needs to be uploaded to the server, and the entire process of interaction between the Athlete and the AI ​​occurs in real time. This way, our AI will be able to process videos directly on users' devices, reducing the wait time between completing exercises and receiving recommendations.

# Labelling tools

Automatic labeling algorithm, ALA (centralized). ALA helps integrate CV model results into Sport Knowledge LLM for real-time reasoning and feedback to athletes during sports training. Stack: Voxel51. Will be deprecated soon as well.

# Sport Knowledge LLM

  • Core of fine-tuned sports coaching agents/transformers that allows you to train people in any sport. Stack: Voxel GPT=>Llama
  • Vector database (centralized). A fast in-memory vector database for LLM and AIGC students in Sports Knowledge. Stack: Epsilla

# Gamification agents

  • Platform for delivering AIGC and fine-tuning coaching models. Stack: Voxel51 tools.

# Reasoning agents

  • A special AI models that can recognize and understand the quality of sports training results and reward athletes with $ACTI. Stack: Voxel GPT, migrating to Llama.

# Web3 tools

  • Web3 frontend/backend dashboard allows to manage $ACTI tokens, bridge it across the connected ecosystems, stake tokens for profit, vote in DAO and many other functions.
  • Bridges (ICP⇔BSC, SOL, TON, SKALE);
  • Oracles (Decentralized Data Storage⇔Sport Knowledge LLM);
  • Actiq DAO (decentralized). The DAO based on ICP (Internet Computer Protocol) subnet allows the community to take control of the whole project development, incentivizing protocol and monetizing of AI models.