Will u lose interest in GO if AI is better than people?

[goTony]

Hello all,

I find GO fascinating, due to many of its different aspects. I find it amazing that AI has not defeated the top players already. And that is one of the beauties of it for me. Will any of you lose interest, or will some of the luster of GO be diminished when AI has outstripped all of us?

[DrStraw]

I'll be dead so it won't make any difference to me.

[Kirby]

Computers are already better than me. It doesn't make much difference to me if they are a little better than me or a lot better than me.

[Krama]

If they get to the level of chess engines I will be even more interested in go than I am today.

[SaiLens]

If chess is any indication, people won't really care.

I wonder. If fascination is based on intractability, that fascination should vanish as soon as the problem is solved.
But sometimes, fascination arises from the meta-game. Some people still play tictactoe to pass the time and you don't even need a computer to come to the conclusion that there can't be a winner.

I'm just waiting for the return of jigo. To play a game with so much freedom and to end it with an even score... that fascinates me.

[Jhyn]

Computers are already better than me. It doesn't make much difference to me if they are a little better than me or a lot better than me.

I'm just waiting for the return of jigo. To play a game with so much freedom and to end it with an even score... that fascinates me.

I agree with these opinions.
Althought I would like to be stronger than AI at least once before it definitely overtakes me, and I am in favor of using integer komi regardless of AI strength.

[CnP]

It wouldn't bother me. It could lead to high quality automatic game analysis, or the systematic exploration of billions of variations of game-space. I'd love to see what sort of a game two 30-Dan AI's play. To me it's a bit like saying since cameras can take perfect pictures why bother painting a picture (well maybe not but I like the analogy). The fact that computers have been relatively weak at Go has never been much of an interest for me though. Chess programs have reached that level now but it doesn't seem to have killed the game.

[Calvin Clark]

The short answer, for me, is no. However...

I guess it depends on the form such superiority takes. If large swaths of opening options are demonstrated to be clearly inferior, it will be very interesting to see which ones, but for me it will spoil a lot the fun of playing the game. Today, if a pro says, "no one plays this any more because it's thought to be good for white" I take it with a grain of salt, because I know pros are not immune to fashion, or confirmation bias, or group myopia. Also, they can't quantify the level of perceived inferiority.

I don't play chess, but I know that many chess players use computers to help analyze their own games and it would be interesting to see that become possible in Go. Of course, to improve you have to think for yourself rather than just blindly accepting what the computer tells you, but that's no different than with human teachers. Anyone---human or computer---can give you a better move, but understanding how it is better and truly utilizing that knowledge in future games is another thing altogether. Is it better as long as I can find the pro-level continuation but inferior otherwise? [Incidentally, when a teacher says "this is a strong move" that's what they really mean. It means "good for me, not for you." ] Is this slack move the computer favors only the best because I am slightly ahead and it maximizes winning probability, but only if I can keep perfect concentration for the rest of the game, which I almost never do?

There is also the question of whether I am playing against a cyborg when playing online. This is unsettling, but I think got over it long ago, when I learned that many players in Asia play in shared kiosks and may have kibitzers helping them, so you never know who is on the other end and what they are doing anyway.

Anyway, if monkeys start becoming better than go than humans, I will become less interested in playing go but more interested in monkeys. Same for AI.

[jdl]

The short answer, for me, is no. However...

I guess it depends on the form such superiority takes. If large swaths of opening options are demonstrated to be clearly inferior, it will be very interesting to see which ones, but for me it will spoil a lot the fun of playing the game. Today, if a pro says, "no one plays this any more because it's thought to be good for white" I take it with a grain of salt, because I know pros are not immune to fashion, or confirmation bias, or group myopia. Also, they can't quantify the level of perceived inferiority.

This is related to the nagging worry I have about go AI's. If they lead to an opening book like chess, then I'm definitely less interested in the game. If the computers simply whip pros and help with game analysis, then it doesn't affect me at all.

[Pio2001]

Is it better as long as I can find the pro-level continuation but inferior otherwise?

This is indeed a problem with the game of Othello, where AI are much stronger than humans.
With a friend, we had each of our games analyzed by the software, and that helped us a lot, granted that the search depth of the software was limited to a level much stronger than ourselves, but not too much stronger.

Otherwise, the software begin to play what we call "inhuman" moves. I once wrote an article about one of such moves, because it clearly appeared to be the worst possible thing to do in the game according to the basics of Othello. A bit like suiciding a potential of 50 points in a go game.
It was possible to explain the move after a careful exploration of all the variations proposed by the computer. But there was nothing to learn from it. It was just a boring enumeration of variations and variations, all leading to what would have been, say, a massive reduction of the 25 intersections taken by the opponent, against all odds. The AI just read deeper.

[skydyr]

Is it better as long as I can find the pro-level continuation but inferior otherwise?

This is indeed a problem with the game of Othello, where AI are much stronger than humans.
With a friend, we had each of our games analyzed by the software, and that helped us a lot, granted that the search depth of the software was limited to a level much stronger than ourselves, but not too much stronger.

Otherwise, the software begin to play what we call "inhuman" moves. I once wrote an article about one of such moves, because it clearly appeared to be the worst possible thing to do in the game according to the basics of Othello. A bit like suiciding a potential of 50 points in a go game.
It was possible to explain the move after a careful exploration of all the variations proposed by the computer. But there was nothing to learn from it. It was just a boring enumeration of variations and variations, all leading to what would have been, say, a massive reduction of the 25 intersections taken by the opponent, against all odds. The AI just read deeper.

The same can be true of go now. You can tell, say, a 15 kyu that this cut works because 15 moves later white's corner loses the capturing race, but it's meaningless when their reading horizon is nowhere near that. A move can be an overplay on some level, but when the punishment leads to an incredibly tense and complex fight that one side wins only by following a specific path, there are too many places to go wrong to make studying it worthwhile, or sometimes even treating it like an overplay.

[Pio2001]

Agreed. By the way, with Othello AI, the most interesting things to learn from the machine occur in mid-game, where the evaluation function is working to choose what looks like the best move according to theory.
Near the endgame, the evaluation becomes mathematically exact, and the AI displays the optimal sequence and gives the exact final score. At this stage, the moves become much less interesting to see, because they don't obey to any strategic principle at all. They are just "the best move in that position". Knowing it doesn't help to find the best one in another game, because in another game, the best move is not the same.

[emeraldemon]

There was a brief, glorious window of time in which I was better at go than any computer. But the programs and hardware kept improving, and I did not. Now my phone is better than me at go. (Champion Go on my LG G3, in case you're wondering). In a way it's nice, I haven't been playing much go lately and my phone is my most frequent opponent. It's always there, I can play a few moves and put it away to think if I want, but it's easy to just play mindless blitz for a few minutes and I worry if maybe that reinforces bad habits.

Whenever I win a game I decrease the handicap, and whenever I lose I increase it, using komi between ranks like so:
2 stones 6.5 komi <-> 2 stones 0.5 komi <-> 3 stones 6.5 <-> 3 stones 0.5 <-> etc.
I have occasionally beaten it down to black no komi, but usually hover between 2 and 3 stones, and I've gone on losing streaks all the way to 5 stones. If I ever improve to the point of taking white, I will probably get bored with it, but it seems more likely that a new version will come out, or I will upgrade my phone, and it will start pushing me to 6 and 7 stones...

In case you were curious, here's a recent 2 stone game I lost.

(
;FF[1]SZ[19]PB[Player]PW[COMLv10]BS[0]WS[10]KM[6.5]HA[2]RU[JP]AP[Champion Go]VW[]
GN[Champion Go]GC[]DT[2015-12-02 21:50:17]RE[W+2.5]
AB[pd][dp]
;W[qp]TL[0,0];B[cd]TL[3,0];W[ed]TL[0,0];B[ec]TL[9,0]
;W[fc]TL[0,0];B[dc]TL[1,0];W[fd]TL[0,0];B[cf]TL[1,0]
;W[jc]TL[0,0];B[op]TL[1,0];W[lp]TL[0,0];B[pm]TL[4,0]
;W[oo]TL[0,0];B[no]TL[1,0];W[on]TL[0,0];B[pp]TL[0,0]
;W[qn]TL[0,0];B[qq]TL[3,0];W[qo]TL[0,0];B[nn]TL[1,0]
;W[om]TL[0,0];B[pr]TL[1,0];W[qj]TL[1,0];B[jq]TL[39,0]
;W[qf]TL[1,0];B[nc]TL[4,0];W[rd]TL[1,0];B[qc]TL[1,0]
;W[of]TL[1,0];B[ck]TL[18,0];W[fq]TL[1,0];B[hq]TL[6,0]
;W[cq]TL[1,0];B[cp]TL[1,0];W[dq]TL[1,0];B[ep]TL[1,0]
;W[gq]TL[1,0];B[hp]TL[5,0];W[fo]TL[1,0];B[fp]TL[1,0]
;W[gp]TL[1,0];B[go]TL[1,0];W[ho]TL[1,0];B[gn]TL[1,0]
;W[hn]TL[1,0];B[fn]TL[2,0];W[kq]TL[1,0];B[jp]TL[15,0]
;W[ln]TL[1,0];B[mq]TL[9,0];W[jo]TL[0,0];B[lq]TL[4,0]
;W[kp]TL[1,0];B[kr]TL[1,0];W[hl]TL[1,0];B[nm]TL[19,0]
;W[nl]TL[0,0];B[mm]TL[60,0];W[lk]TL[0,0];B[fk]TL[94,0]
;W[fl]TL[0,0];B[el]TL[14,0];W[gk]TL[1,0];B[fm]TL[3,0]
;W[fj]TL[1,0];B[gl]TL[1,0];W[hj]TL[1,0];B[ej]TL[2,0]
;W[fi]TL[0,0];B[lc]TL[4,0];W[ei]TL[1,0];B[mf]TL[2,0]
;W[dj]TL[1,0];B[dk]TL[1,0];W[ek]TL[1,0];B[hc]TL[8,0]
;W[hd]TL[1,0];B[ic]TL[29,0];W[id]TL[1,0];B[jb]TL[1,0]
;W[kc]TL[0,0];B[kb]TL[1,0];W[ke]TL[1,0];B[ld]TL[2,0]
;W[lg]TL[1,0];B[mg]TL[2,0];W[lh]TL[1,0];B[mh]TL[24,0]
;W[mi]TL[1,0];B[oh]TL[2,0];W[ni]TL[1,0];B[pg]TL[2,0]
;W[pf]TL[1,0];B[og]TL[4,0];W[qg]TL[0,0];B[qh]TL[2,0]
;W[rh]TL[0,0];B[qi]TL[1,0];W[ri]TL[1,0];B[cj]TL[13,0]
;W[ci]TL[1,0];B[ej]TL[1,0];W[bp]TL[1,0];B[bo]TL[2,0]
;W[bq]TL[1,0];B[cn]TL[3,0];W[di]TL[1,0];B[ef]TL[18,0]
;W[ff]TL[1,0];B[eg]TL[5,0];W[fg]TL[1,0];B[ll]TL[8,0]
;W[kl]TL[0,0];B[jn]TL[23,0];W[io]TL[1,0];B[kn]TL[4,0]
;W[ko]TL[1,0];B[lm]TL[3,0];W[hr]TL[0,0];B[ir]TL[6,0]
;W[ip]TL[0,0];B[iq]TL[1,0];W[is]TL[1,0];B[jr]TL[2,0]
;W[gr]TL[0,0];B[kk]TL[6,0];W[jl]TL[0,0];B[mk]TL[4,0]
;W[lj]TL[0,0];B[nk]TL[1,0];W[ok]TL[0,0];B[km]TL[2,0]
;W[pi]TL[1,0];B[ph]TL[1,0];W[oi]TL[1,0];B[rq]TL[2,0]
;W[mp]TL[0,0];B[np]TL[2,0];W[hm]TL[0,0];B[eq]TL[4,0]
;W[er]TL[0,0];B[sn]TL[6,0];W[so]TL[0,0];B[ro]TL[1,0]
;W[sm]TL[1,0];B[sp]TL[3,0];W[rn]TL[1,0];B[so]TL[1,0]
;W[rp]TL[0,0];B[sq]TL[1,0];W[rl]TL[1,0];B[rc]TL[2,0]
;W[od]TL[1,0];B[oc]TL[1,0];W[nd]TL[0,0];B[md]TL[7,0]
;W[nh]TL[0,0];B[ng]TL[2,0];W[qd]TL[0,0];B[lf]TL[64,0]
;W[kf]TL[0,0];B[sd]TL[23,0];W[re]TL[0,0];B[gb]TL[60,0]
;W[fb]TL[0,0];B[eb]TL[1,0];W[ea]TL[0,0];B[da]TL[3,0]
;W[fa]TL[1,0];B[bg]TL[7,0];W[bh]TL[0,0];B[ee]TL[6,0]
;W[dd]TL[0,0];B[cb]TL[3,0];W[de]TL[0,0];B[df]TL[2,0]
;W[ce]TL[1,0];B[be]TL[1,0];W[bj]TL[0,0];B[bk]TL[1,0]
;W[fe]TL[0,0];B[ga]TL[2,0];W[gc]TL[1,0];B[hb]TL[2,0]
;W[se]TL[1,0];B[sc]TL[2,0];W[pe]TL[1,0];B[aj]TL[2,0]
;W[pc]TL[1,0];B[pb]TL[1,0];W[bi]TL[0,0];B[po]TL[29,0]
;W[pn]TL[0,0];B[jm]TL[4,0];W[il]TL[0,0];B[kd]TL[3,0]
;W[jd]TL[0,0];B[eh]TL[2,0];W[le]TL[0,0];B[me]TL[1,0]
;W[pd]TL[0,0];B[fh]TL[2,0];W[gh]TL[0,0];B[nj]TL[40,0]
;W[oj]TL[0,0];B[ek]TL[10,0];W[gm]TL[0,0];B[fl]TL[1,0]
;W[ai]TL[0,0];B[ak]TL[1,0];W[ag]TL[0,0];B[af]TL[1,0]
;W[bd]TL[0,0];B[cc]TL[2,0];W[cg]TL[0,0];B[ah]TL[3,0]
;W[dg]TL[0,0];B[bc]TL[5,0];W[ag]TL[0,0];B[bf]TL[6,0]
;W[dh]TL[0,0];B[ah]TL[3,0];W[ad]TL[0,0];B[ac]TL[1,0]
;W[ag]TL[0,0];B[ae]TL[1,0];W[ao]TL[0,0];B[an]TL[1,0]
;W[ap]TL[0,0];B[ne]TL[8,0];W[qb]TL[0,0];B[rb]TL[1,0]
;W[ob]TL[0,0];B[qa]TL[1,0];W[nb]TL[0,0];B[mc]TL[1,0]
;W[ia]TL[0,0];B[ib]TL[5,0];W[mb]TL[0,0];B[lb]TL[1,0]
;W[pl]TL[0,0];B[ml]TL[6,0];W[ol]TL[0,0];B[lr]TL[5,0]
;W[hs]TL[0,0];B[js]TL[1,0];W[im]TL[0,0];B[ah]TL[11,0]
;W[mo]TL[0,0];B[mn]TL[1,0];W[ag]TL[0,0];B[lo]TL[2,0]
;W[bd]TL[0,0];B[ad]TL[1,0];W[ln]TL[0,0];B[ah]TL[6,0]
;W[bm]TL[0,0];B[bn]TL[2,0];W[ag]TL[0,0];B[lo]TL[1,0]
;W[kj]TL[0,0];B[ln]TL[3,0];W[mj]TL[0,0];B[ah]TL[4,0]
;W[jg]TL[0,0];B[ag]TL[2,0];W[in]TL[0,0];B[tt]TL[8,0]
;W[nf]TL[0,0];B[oe]TL[7,0];W[oa]TL[0,0];B[ma]TL[2,0]
;W[tt]TL[0,0];B[tt]TL[0,0]
)

[Anzu]

Chess computers are so strong, they only take one second per move to beat me. I look forward to having an ultra-fast Go engine in the future. Why play humans when your Go software moves instantly?

[horace]

No, I won't.

Great AI will become a good tool as a personal reviewer. So I can get advice from the machine when I don't know how to handle a position ...