GatewayRedbirds.com

First we had Microsoft's AI Twitter account rapidly turn into a sociopath, and now this:

Elite's AI Created Super Weapons and Started Hunting Players. Skynet is Here.

It all started after Frontier released the 2.1 Engineers update. The release improved the game's AI, making the higher ranked NPCs that would fly around Elite's galaxy more formidable foes. As well as improving their competence in dog fights, the update allowed the AI to use interdiction hardware to pull players travelling at jump speed into normal space. The AI also had access to one of 2.1's big features: crafting.

These three things combined made the AI a significant threat to players. They were better in fights, could pull unwary jump travellers into a brawl, and they could attack them with upgraded weapons.

There was something else going on, though. The AI was crafting super weapons that the designers had never intended.

Players would be pulled into fights against ships armed with ridiculous weapons that would cut them to pieces. "It appears that the unusual weapons attacks were caused by some form of networking issue which allowed the NPC AI to merge weapon stats and abilities," according to a post written by Frontier community manager Zac Antonaci. "Meaning that all new and never before seen (sometimes devastating) weapons were created, such as a rail gun with the fire rate of a pulse laser. These appear to have been compounded by the additional stats and abilities of the engineers weaponry."

Teaching Robots to Feel: Emoji & Deep Learning

I didn't know whether to put this here, the uber thread, or the Are we Effed thead.

Driverless cars are the next frontier in transportation, and Uber wants to be ready.

The ride-hailing giant is pushing to take advantage of the emerging technology while also trying to build those cars itself

link.

New artificial intelligence
beats tactical experts in combat simulation

Artificial intelligence recently won out during simulated aerial combat against U.S. expert tacticians. Importantly, it did so using no more than the processing power available in a tiny, affordable computer (Raspberry Pi) that retails for as little as $35.

One Giant Step for a Chess-Playing Machine

In early December, researchers at DeepMind, the artificial-intelligence company owned by Google’s parent corporation, Alphabet Inc., filed a dispatch from the frontiers of chess.

A year earlier, on Dec. 5, 2017, the team had stunned the chess world with its announcement of AlphaZero, a machine-learning algorithm that had mastered not only chess but shogi, or Japanese chess, and Go. The algorithm started with no knowledge of the games beyond their basic rules. It then played against itself millions of times and learned from its mistakes. In a matter of hours, the algorithm became the best player, human or computer, the world has ever seen.

Suppose that deeper patterns exist to be discovered — in the ways genes are regulated or cancer progresses; in the orchestration of the immune system; in the dance of subatomic particles. And suppose that these patterns can be predicted, but only by an intelligence far superior to ours. If AlphaInfinity could identify and understand them, it would seem to us like an oracle.

We would sit at its feet and listen intently. We would not understand why the oracle was always right, but we could check its calculations and predictions against experiments and observations, and confirm its revelations. Science, that signal human endeavor, would reduce our role to that of spectators, gaping in wonder and confusion.

Maybe eventually our lack of insight would no longer bother us. After all, AlphaInfinity could cure all our diseases, solve all our scientific problems and make all our other intellectual trains run on time. We did pretty well without much insight for the first 300,000 years or so of our existence as Homo sapiens. And we’ll have no shortage of memory: we will recall with pride the golden era of human insight, this glorious interlude, a few thousand years long, between our uncomprehending past and our incomprehensible future.

I, for one, bow to our new robot overlords.

Just a quick note as I've been watching the DeepMind games a little bit and specifically the chess games...

But, with chess (and I'm no expert) it seems that back in the 1990s supercomputers got a point where they could beat the best human chess players in the world pretty consistently. And, with advances in technology/computing power, they've only gotten better.

Back in December of 2017, iirc, AlphaGo/DeepMind competed against Stockfish 8 (top engine but not the latest version) in 100 games winning 27 or so and drawing the rest. Even the best human chess players would probably be lucky to win one game out of 100, probably one game out of 1000. So, this was quite surprising.

Since the 1990s, chess has, imo, evolved to copy what the engines do which is, more or less, incrementally improve positions taking a pawn here or there until it overwhelms an opponent. And, for supercomputers, this is easy to do as they can assess and project multiple lines multiple moves in advance and have a numerical value for each position.

To this day, top players will prepare for tournaments/games using Stockfish to analyze positions and try to see what opponents will do/play and how to attack those moves. If you watched any of the world chess championship, you'd note that Magnus Carlsen beat Fabiano Caruana but neither player won a single game in the classical time format which just goes to show how automated the best players in the world have become. Compared to 50 years ago, it's just, again imo, memorizing openings/moves and not letting an opponent get an attack going.

While I'm not sure how exactly DeepMind calculates positions other than obviously having access to all the previous games it played, but watching how it beat Stockfish was truly incredible because we know how Stockfish calculates positions. As a matter of fact, there was one game released last year that shows DeepMind sacraficed a rook and Stockfish calculated the position as completely winning for Stockfish. But, within 5 or 6 moves, the Stockfish had no good moves to make. Eventually Stockfish tried to sacrifice it's queen to get out of the quagmire DeepMind had it in but it didn't matter. DeepMind won the game. And, overall, DeepMind has shown an ability to make sacrifices and play moves no computer would ever think to do because of the temporary loss of material.

My understanding is very limited so take this fwiw, but from my understanding, an engine like stockfish takes a position and then calculates which moves increase the score or position the most for that position (lets call these the 'next moves') and then looks further into the game at how many different lines for those 'next moves' would go and can pick the best one. But, it is reliant on calculating the correct 'next move' to proceed correctly. And engines have become very good at doing this.

But, it doesn't appear that DeepMind is taking that same approach where it calculates the position after the 'next move' and proceeds from there. It looks more and more like it is capable of taking a more holistic approach to each move while still being able to do deep calculations well into the future for the game.

Here's a link to one of the games released back in 2017.



All in all, really amazing stuff.

Edited to say: Obviously, humans have been working on chess engines and their algorithms for decades. DeepMind got the rules and then played itself for a few hours and in that minuscule, relatively speaking, time frame became better than anything humans created over the past 4ish+ decades.

the limitations come when we address a problems that are not well-defined in terms of inputs and outputs. With problems like chess, you have your simple rules. when you have a scientific problem like protein folding, you have a small set of amino acids and the laws of physics. when you have a complicated system like cancer, the calculations become tremendous, so I'm skeptical we are going to have that level of AI in our lifetime. There is definitely an advantage to trying to piece together AI's logic for simple problems but the logic is not so straightforward. Still exciting.

Great post AW. It’s why I really think Chess960 may offer a future for human players and removes so much automaton like play.