{"id":12762,"date":"2023-07-04T12:46:22","date_gmt":"2023-07-04T12:46:22","guid":{"rendered":"https:\/\/aurelis.org\/blog\/?p=12762"},"modified":"2023-07-04T14:20:47","modified_gmt":"2023-07-04T14:20:47","slug":"from-tool-to-autonomy","status":"publish","type":"post","link":"https:\/\/aurelis.org\/blog\/artifical-intelligence\/from-tool-to-autonomy","title":{"rendered":"From Tool to Autonomy"},"content":{"rendered":"\n

When can a tool \u2013 gradually \u2013 become an autonomous agent, and how must we deal with this?<\/h3>\n\n\n\n

What is an agent?<\/a><\/p><\/blockquote>\n\n\n\n

And what is a tool? For instance, is your body (or your left hand) a tool of your brain? Is your entire body \u2013 including your brain \u2013 a tool of your mind? Or does it all depend on your perspective? Then where is this perspective taking you? Conversely, can your diary (in your smartphone) be seen as part of your extended memory \u2015 therefore, you?<\/p><\/blockquote>\n\n\n\n

We attribute goals to instrumental tools.<\/strong><\/p>\n\n\n\n

Until now, people have done so metaphorically. For instance, the goal of the hammer is to hit the nail<\/a>. With complicated tools, the metaphorical goals may become more sophisticated.<\/p>\n\n\n\n

With complex<\/a> tools, however, autonomy starts creeping in, and the metaphorical stance becomes less straightforward. When do goals become the goals of the tools themselves? In other words, when does viewing \u2018autonomous tools\u2019 as mere tools an intolerable case of reductionism<\/a>? With increasingly complex super-A.I., this issue will gradually become more pertinent.<\/p>\n\n\n\n

Since there is no discrete border between both situations, we will need to consider all kinds of fogginess. That\u2019s OK if we find more abstract principles to guide us. Working case by case will not do eventually. It will create a big mess instead.<\/p>\n\n\n\n

As in many cases of complexity, we need fundamental insights.<\/p>\n\n\n\n

Anthropomorphism and its counterpart<\/strong><\/p>\n\n\n\n

Readily attributing goals to tools is an example of anthropomorphism. Even knowing that it\u2019s metaphorical, we may emotionally react otherwise, for instance, towards robots with human appearance \u2015 nothing fundamentally wrong with that.<\/p>\n\n\n\n

The counterpart is also possible: not seeing the likeness with us for thinking that something different cannot principally be human-like. This may be the fate of super-A.I. in the eyes of many people, even for still quite a while.<\/p>\n\n\n\n

We will need a balance between both stances. History shows we\u2019re not good at such balancing acts, even concerning our fellow human beings. This goes from embryos to mass extinctions of fellow humans who are apparently not genuinely human \u2015 like some philosophical zombies<\/a> or worse.<\/p>\n\n\n\n

With super-A.I., we must be cautious because of obvious reasons.<\/p>\n\n\n\n

The law<\/strong><\/p>\n\n\n\n

A super-challenging question will be the judicial implication of Artificial Consciousness<\/a> in two directions:<\/p>\n\n\n\n