++ We have taken down all the Kindle booklets on 31 Dec 2021 in order to pursue a different path, with software coming first. The booklets will return at a later stage!
Tailored content can be provided upon request. Submit your technical specification through the contact form in the About page at https://www.tenproblems.com/about/ and get your quote.
Literature Review: Computer Games problem for Artificial Intelligence
This “Ten Problems for Artificial Intelligence in the 2020s” booklet identifies ten relevant areas from very recent contributions put forward at academic level in the form journal articles, conference proceedings and students theses. Ten freely accessible internet references have been selected for each area and direct links are provided at the end of each chapter for own consultation. Our selected references do not intend to mirror ranking indexes nor establish novel classifications. On the contrary, they are meant to represent peer-reviewed, diverse and scientifically-sound case studies for vertical dissemination aimed at non-specialist readers. They will also be able to scoop even more references through the bibliography that is reported at the end of each selected reference.
Without further ado, these are the ten problems that we are going to introduce in this booklet:
- computer games,
- intelligent optimization,
- human intervention,
- ethical issues,
- software development,
- open source,
- operations research.
Each problem has its own dedicated chapter made of an introductory section, a short presentation of the ten selected references and a conclusions section.
The final chapter of this booklet will report the conclusions from each chapter again in order to provide a complete executive summary.
1 Computer Games
THE PROBLEM — Computer games have been historically the easiest (without real-world implications when deployed) and the most approachable (because of the sheer size of the market and the sustained competition) industry for Artificial Intelligence (AI) research, development. In the recent years, though, the search for Artificial General Intelligence (AGI) and the deployment to real-world and business, social and governmental applications has demoted computer games to the more marginal role of one virtual laboratory among others.
CASE STUDIES — … buy this booklet from Amazon …
CONCLUSIONS — A new formal definition of intelligence based on Algorithmic Information Theory, describing intelligence as skill-acquisition efficiency, must be considered instead of single-task optimization. Beating the player while behaving as human-like as possible and taking care of multiple opponents may not be possible simultaneously. It is still difficult to apply Deep Reinforcement Learning to certain real-world problems because each action is not predictable, and we cannot know why the results are coming out. Team sports occupy a special, possibly unique, niche in the AI development and benchmarking landscape. Tiny neural networks can be evolved to decide actions based on the encoded observations, to achieve results comparable with the deep neural networks. While behavioral cloning is not able to reach the performance of a human player, it is still able to learn the basic mechanics of most games. While deep reinforcement learning agents perform well most of the time, the question of whether unsafe behavior may occur in corner cases is an open problem. Intrinsic motivation offers a method to augment these algorithms with social preferences, echoing the social cognitive processes observed in humans. It seems like OpenAI is becoming a marketing proxy for Microsoft’s Azure cloud and will help spot AI startups that might qualify for acquisition by Microsoft in the future. Cooperative multi-agent behavior, the capacity of an agent to observe a varying number of entities, and establishing a single model to solve several tasks are now available in Unity ML-Agents.
TEN FREE REFERENCES FROM THE INTERNET — … buy this booklet from Amazon …
booklet updated on 19 Jun 2021, now on sale as version 1.2