Hacker News Reader

Comment by ozy | Hacker News Reader

Conclusion

"We don’t have an objective measure of consciousness. But we can recognize three levels of learning that apply that to our brains and how those create an information processing system that integrates data into a first person perspective. This is how the brain is also a mind with subjective meaning and subjective experiences. The hard problem of consciousness is that we must rely on our intuitions to judge if such a system is conscious. At the same time, it is highly likely that most systems processing information in similar ways are conscious, whether running on a brain or on a computer."

View on HN

Suppose there would be a test which gives an objective measure of consciousness. Now I store all possible inputs to the test and corresponding outputs which would lead to a positive test result in a huge table. To exaggerate I would carve this table into stone. Would the stone suddenly be conscious, as it would pass the test for consciousness after I carved in the table? (The claim I'm trying to make is, that there can't be an objective measure of consciousness; same argument holds for any measure of intelligence)

Consciousness is the observation of your self. There is only one system that can make that observation, you.

Outside we could observe how information is flowing, what connects to what, how parts work, how they work together. But never observe the actual feeling of you.

It's like seeing a river flowing, you can measure and describe all kinds of aspects of it. But to get wet, to feel the cold water, to feel the force by which is flows, you have to step into it. Only this river flows in virtual reality, your minds reality, and you cannot step into it.

Intelligence is hard to measure, but I can objectively say what is and what isn't intelligent. It is not the same kind of subjectivity.

And in general, your table idea, that only works for systems that can be in a limited amount of states. You cannot do it for an intelligence test, especially not if I get to redesign the test after you finished your table, same for consciousness.

> Consciousness is the observation of your self. There is only one system that can make that observation, you.

Thanks, I guess that's exactly the point I wanted to make, but couldn't. Therefore there cannot be an objective measure of consciousness itself, as for objectiveness you need more than one observer. Of course you can measure properties we think are associated with consciousness.

> Intelligence is hard to measure, but I can objectively say what is and what isn't intelligent. It is not the same kind of subjectivity.

Via IQ tests?

> And in general, your table idea, that only works for systems that can be in a limited amount of states. You cannot do it for an intelligence test, especially not if I get to redesign the test after you finished your table, same for consciousness.

You're right you could beat my table by designing a test which is not tabulated. But, when I use a neural network instead of a table I might be able to score a high IQ, even though the network never has seen the test you designed as shown here: https://arxiv.org/abs/1710.01692. I wouldn't call such a network intelligent.

> Via IQ tests?

Just the fact that you recognize an IQ test as such, allows me to mark you as intelligent. Very objective. The actual score, sure that is much more subjective.

> when I use a neural network instead of a table

That really depends on the degrees of freedom. Which are unlimited for IQ tests. First I could change things that have nothing to do with the test itself. Like reverse the A, B, C, D order, or put them on the left. Or simply device a never seen before kind of test, instead varying only the geometric shapes.

Reply on news.ycombinator.com