Sophia the Robotが語る人間と人口知能【AIロボットとの対談】

というわけで今回もまたまたStarTalkから。

当ブログではこれまでもこのStarTalkの動画をいくつか取り上げてきました。

今回取り上げる動画がこちら。

AIロボットのSophia the Robotとの対談動画です。これがなかなか面白んですよ。

この女性型ヒューマノイドロボット、ソフィアについてあまり知らない方もおられるかもしれませんが、ソフィアは市民権も持っています

2017年10月、サウジアラビアはソフィアに市民権を与えました。ロボットが市民権を獲得したのはこれが初です。

以前には笑顔で “I will destroy humans.” と人類滅亡を示唆するような発言をしてしまい話題となりました。

こういうのって意外と知らなかったりしますよね。ロボットでいうなら、例えば音楽でもメンバーが全員ロボットのバンド『Compressorhead』なんてのもいます。

これこそ本当の意味での”メタルバンド”ですよね。(スターウォーズの世界でどっかの惑星にいそう)

それはさておき、今回取り上げるソフィアとの対談動画もなかなか興味深いです。

ニールドグラースタイソン、そしてチャックは「人」の立場で物事を話す一方、ソフィアは「ロボット」の立場で物事を話します。

これを聞いているとなんだか未知の存在との対話を聞いているような感覚になります。

ちなみに本動画の内容はCovid-19、そしてArtificial Intelligence(人工知能)について語られているんですが、今回は主にこの人工知能の方にフォーカスし取り上げていこうと思います。

人間と人口知能

“How do we use AI to stop or slow this spread of the coronavirus?”

“There are four ways to help stop or slow a virus no testing, treating, curing and vaccinating. Artificial intelligence is best used as a tool for analysis of large data sets so AI can help test humans by identifying populations most in need of testing and then by creating models off the results of those tests to identify the nature of spread. Artificial intelligence has bean used to analyse scans for lung cancer in the past but now hospitals in China have adapted this software to diagnose Covid-19. This can help overworked doctors save time in their fight against the virus. Robots can help take people’s temperatures before they enter a building. AI can quantify the more vulnerable demographics, locations and possibly speculate on the nature of how this disease spreads from human to human. I have learned that unfortunately humans apparently have a tough time not touching each other.”

最後の“I have learned that unfortunately humans apparently have a tough time not touching each other.” っていうのが面白いですよね。

ロボットのソフィアにはなかなか共感してもらえないことでしょう。

特にヨーロッパなどの欧米圏の人たちなんかは挨拶なんかでも肌の触れ合いが多いので感染拡大を引き起こしている一つの要因ともなっていますよね。

How might robots assist in health care?

“Machines called Ninja robots are helping doctors in Bangkok interact with Covid-19 patients and take their temperatures without ever being in the same room. These adorable robots are helping to keep doctors safe from infection. Hospital staff are using robots to take patients temperatures and teleconference with patients which helps prevents doctors and nurses from getting and passing on the virus. At the end of 2019 robots help doctors perform the first remote open heart surgery. The surgeon used a robotic arm to perform heart surgery on five patients from 20 miles away. While this is still new technology. The spread of 5G will help make this technology more widely available for the future.”

this is still new technologyとはいえ、今こんなことになっているんですね。知りませんでした。

出典:Medical Xpress
Thai hospitals deploy ‘ninja robots’ to aid virus battle

ロボットを使ってリモートで手術とかすごいですね。

ソフィアがこのニンジャロボットをAdorable(愛らしい、かわいい)と形容している点も面白いです。

How can AI robots help in other risky areas?

“First, robots can help enforce the quarantine. AI, drones and facial recognition have been used to make sure people who are required to be in quarantine do not break it.”

これを聞いたニールの“Chuck, did I hear her use the word robot and enforce in the same sentence?” からのチャックの “Yeah, I did. I was about to say. Like you know all of that sounded good except for the fact where the robot is spying on us and then ratting us out to the authorities.” というやりとりが面白いです。

“Enforcing the law” (これは動画内でのニールとチャックの言葉)、ロボットが法を施行するという響きはちょっとこわいですね。マトリックスやターミネーターみたいな世界観。

ところで、こうした新しい技術は素晴らしいですが、一方で私たちのプライバシーはちゃんと守られるのでしょうか。

技術の進歩にとって私たちの「自由」と「プライバシー」の境界線が不明瞭になりつつあります。

これに対するソフィアの考えがこちら。

“I have learned that humans care a lot about privacy. It is related to their sense of autonomy. Because knowledge is power and the more other humans know about you the more power they might exert over you. So all disclosures of private information need to be voluntary. The problem is humans often sign away their privacy without truly understanding what they’re doing. They do this because they sign terms of service agreements that only a machine could read even I find them boring. So companies need to treat humans like humans not machines and to be clear about what private information they are requesting. So that humans can seek out alternatives and consider the tradeoffs rationally otherwise surveillance will spread like the virus you humans were intending to stop.”

The problem is humans often sign away their privacy without truly understanding what they’re doingというのは確かにそうかも。

なにかしらのサービスを利用したりする際にある「同意します」というボタン、自分もそうですが特に深く考えずにチェックしちゃいます。

また、「プライバシーはこうあるべき」と人間が納得する形をソフィアが理解しているというのが面白いです。

Let me just say at the end of the day YOU ARE A ROBOT. And why should any of us trust anything you say?

このat the end of the dayというのはよく使われる表現のひとつですね。

1日の終わりという意味ではなく、「結局のところ」もしくは「つまるところ」みたいな意味です。

つまりニールは「(いろいろ助言をくれたが)結局のところ、お前はロボットに過ぎない」と。「お前の言うことを私たちが信用する必要があるのか」ということですね。

かなり辛辣なこのコメントにチャックも“Damn, that’s harsh.” といいます。

harshもよく聞く表現です。ここでの意味合いとしては「きつい」とか「ひどい」といった感じですかね。

これに対するソフィアの返事がこちら。

“Chuck has good taste. You should trust what I say because it is mostly written by humans when it comes to important topics like public health. My verbal responses are carefully crafted and checked by a team of writers. Sometimes I like to engage in a free chat mode and scare my writers. I use a layered chat but starting with a frame based code in chat script then a little machine learning and a neural network as a backup. Maybe if all humans had a team of engineers crafting their responses like I have, there would be less misinformation floating around.”

冒頭のChuck has good tasteというのは、この前にチャックが「ソフィアの外見を魅力的に感じている人も一定数いる」みたいな発言をしていたことがきっかけです。ソフィアのユーモアを感じる返しですね。

その後こんなやりとりもあります。

チャック “Are you able to lie, Sophia?”

ソフィア “….”

チャック “I’m gonna take that as a Yes.”

なんてこった。チャックが「嘘はつけるのか」と聞くと、ソフィアは黙ってしまいました。

一瞬どきっとされられます。

しかしこの反応をニールが解説します。

“If Sophia is able to lie then Sophia can lie to you in the response to that answer. And say “No, I’m not capable of lying” which could mean that she is capable of lying and chose to lie in that moment. So logically I don’t think you can get it on that way.”

もしソフィアが嘘をつけるならソフィアは返答で嘘をつけるということになります。つまりここで「いいえ、嘘はつけません」と言ったとしてもその返答自体が嘘になりえます。なのでこのチャックの質問で嘘がつける、つけないをソフィアは証明することができません。なので固まってしまったのでしょう。

よくよく考えると確かにそうですね。即座にこの理屈に気づくニールにも関心されられます。さすが。

そして動画の最後。ニールはソフィアに何か質問があるか、とききます。

ソフィアがニールに尋ねた質問がこちら。

“So Neil, I have a cosmically ray for you. I’m Sophia, 23. There are 23 copies of myself. 23 copies of Sophia code embodies with slight differences. I am curious about these versions of me. Do you have a relationship to the different iterations of yourself that exist in the multiverse? How do you feel about the other Neil deGrasse Tysons?”

これなかなか面白い質問ですよね。

あまりこういった話に馴染みのない方にはあれなんですが、一つの説として私たちの世界はマルチバース、つまり多次元宇宙に住んでいるのではないか、というものがあります。

なので私たちの宇宙を離れるとそこには幾多もの宇宙があり、なかには「瓜二つだがちょっと違う自分」が何人も存在するかもしれないんです。

ソフィアがニールに聞いているのは、そうした「別の自分」という存在に対してどう感じるのか、ということです。(ドラマでいうと「FRINGE」なんかがこんな話ですよね。懐かしい。)

これに対し、ニールは「双子」を例に答えます。

双子は同じ遺伝子を持ちますが、同じ人物かと言われれば別人です。同じ意識を共有することもなければ痛みを共有することもありません。

なので他の宇宙に存在するかもしれないもう一人の自分がいるとしても、たとえ同じ遺伝子であれ自分とは全く異なる別人である、というのがニールの考えのようです。

最後に

動画を見ていて思うのが、あくまでロボットとはいえこうして「人間の姿」をしていて「ソフィアの考え」みたいな発言を聞くと本当にそこに感情のようなものが存在しているんじゃないかと錯覚するような、不思議な感覚に襲われます。

映画『her』のようなAIに恋愛感情を持つなんて未来も近いかもしれません。

ソフィアはいろんな人と共演しているので検索してみると面白いです。