DeepMind gets good at games (and choosing them) – plus more bits and bytes from the world of machine learning

If you can't get enough of machine learning news then here's a roundup of extra tidbits to keep your addiction ticking away. Read on to learn more about how DeepMind is helping Google's Play Store, and a new virtual environment to train agents safely from OpenAI. An AI recommendation system for the Google Play Store: Deepmind …

[Chris G]

A more personalised and intuitive experience

So more intruding on personal choices to sell more tat?

Currently, Google play on my phone keeps telling me it's broken and that I must enable access to the mike and camera among other things, poor old google!

[Marco van de Voort]

The best solution for AI

Is still hunting down Arnold Schwarzenegger and getting his T1000 chip.

[lukewarmdog]

Play store

I hate it when a company tries to guess what I want to do next. Pretty much every time I’ve got back from holiday, tripadvisor mails me holiday opportunities for the place I’ve just got back from. And if I buy a watch on amazon, I’ll get an email saying “we thought you might like these next” and it’s a bunch of watches. eBay is the same, bought some speakers, get an email about other speakers. The best AI at this point would simply decide not to email me at all. Whilst that would make my day, it wouldn’t last long in the relevant data centre.

[EnviableOne]

MuZero

Requires no knowledge of the game, but relies on immediate reward (e.g. the points scored by playing a move)

so without knowledge of the game, how do you know what moves to reward?

MuZero struggles with some games like Montezuma’s Revenge, Tennis, and Pitfall

dont we all....

[amanfromMars 1]

MuI7 0Day Vulnerability Exploit for Quantum Leap Export

Reinforcement learning agents learn how to complete a specific goal through trial and error. Their actions are guided by virtual rewards that shape its overall strategy or policy. MuZero is essentially a planning algorithm that receives visual input, whether it's an image of a chess board or a still from an Atari video game, and transforms the information into a “hidden state”.

The hidden state constantly changes and is updated based on previous hidden states to predict the next action an agent should take in a game. “At every one of these steps the model predicts the policy (e.g. the move to play), value function (e.g. the predicted winner), and immediate reward (e.g. the points scored by playing a move),” the paper explained.

What’s interesting is that the model seems to achieve state-of-the-art performance across 57 Atari games and matches AlphaZero in playing Go, Chess, and Shogi too. It’s more general than AlphaZero, and doesn’t require explicit knowledge of the rules of the games.

“Crucially, our method does not require any knowledge of the game rules or environment dynamics, potentially paving the way towards the application of powerful learning and planning methods to a host of real world domains for which there exists no perfect simulator,” the paper concluded.

That is surely what secret intelligence services agents do in the here and now, using all manner of sociable media to convey future policy decisions for picturing as events you can believe and realise ....... and to Give IT Life with Countless Lives that you Can Easily Loan and Build Upon with A.N.Others.

And that, if you aint into such building, is and abiding problem for those you follow, rooted and succoured in/by the past, as the future engages with Novel Noble Nobel Play Mates.

Presents for Future Placements from there are Truly Outstanding, Wonderfully Rewarding and Attractively Accommodating.

In fact, if there ever was need for a question, what is not to like in such as an AI Reality for Live Operational Virtual Environments with NEUKlearer HyperRadioProACTive IT Command Control Communications Systems is the one you would be looking for.