Slashdot Mirror
The Lovelace Test Is Better Than the Turing Test At Detecting AI
meghan elizabeth writes If the Turing Test can be fooled by common trickery, it's time to consider we need a new standard. The Lovelace Test is designed to be more rigorous, testing for true machine cognition. An intelligent computer passes the Lovelace Test only if it originates a "program" that it was not engineered to produce. The new program—it could be an idea, a novel, a piece of music, anything—can't be a hardware fluke. The machine's designers must not be able to explain how their original code led to this new program. In short, to pass the Lovelace Test a computer has to create something original, all by itself.
Maybe they mean the "Linda Lovelace" test?
That is all.
When was the last time the average person created something original?
Agreed - there is no reason to require the program be written in perl.
I do recall reading a while back experiments done with AI in which programs compete for resources by generating programs to do tasks given to it (computing sums etc). Some programs did generate code that were completely unexpected.
It raises the question programs that are evolved are designed by the programmer or the program, or the process of evolution. And it also raises the philosophical question about whether we should be more humble and accept that our "creativity" that we think is what makes humans intelligent could be nothing more than a process of the evolution of ideas (I hesitate to use the word meme) that we don't actually originate nor control.
If we consider programs that can create things through evolution as "intelligent", that would ironically make natural selection intelligent, since DNA is a digital program that is evolved into complex things over time that can't be reduced to first principles.
Those who do not learn from commit history are doomed to regress it.
The machine's designers must not be able to explain how their original code led to this new program.
Whoa, whoa, whoa. I have severe problem with this. This is like looking at obscurity and declaring it a soul. The measure of intelligence is that we can't understand it? Intelligence through obfuscation? There should be no way for a designer to not be able to figure out why their machine produced what it did given enough debugging.
It was passed as defined
The Turing Test was not passed, and the only people who claim it was are ignorant reporters looking for an easy story with a catchy headline and tech morons who also believe Kevin Warwick is a cyborg.
The test was rigged in every way possible:
- judges told they were talking to a child
- that doesn't speak English as a primary language
- which was programmed with the express intent of misdirection
- and only "fooled" 30% of the judges.
And, even after all that, Cleverbot did a much better job back in 2011 with a 60% success rate.
This Eugene test outcome was a complete farce -- something to remind everyone that Warwick still exists and to separate the ignorant and sensational tech news trash rags from the more legitimate sources of information.
"What do you despise? By this are you truly known." --Princess Irulan, Manual of Muad'Dib
/)
Oddly computer chess programs may already meet this criteria. The programs usually apply a weight or value to a move and a weight and a value to the consequences down stream of the move. But there are times when the consequences are of equal value at some event horizon and random choices must be applied. As a consequence sequences of moves may be made that no human has ever made and the programmer could not really predict either. As machines have gotten more able the event horizon is at a deeper level. But we might reach the point at which only the player playing white can ever hope to win and the player with black may always lose. We are not in danger of a human ever being able to do that unless we alter his brain.
That's because they keep shifting the goalposts.
They are shifting them again. This new test includes this requirement: The machine's designers must not be able to explain how their original code led to this new program. So now anything we understand is not intelligence??? So if someone figures out how the brain works, and is about to describe its function, then people will no longer be intelligent? Intelligence is a characteristic of behavior. If it behaves intelligently, then it is intelligent. The underlying mechanism should be irrelevant.
Alas, the test that was "passed" was not actually the test Turing proposed.
So it passed the Turingish test.
http://en.wikiquote.org/wiki/I... Detective Del Spooner: Human beings have dreams. Even dogs have dreams, but not you, you are just a machine. An imitation of life. Can a robot write a symphony? Can a robot turn a... canvas into a beautiful masterpiece? Sonny: [With genuine interest] Can you?
That's because they keep shifting the goalposts.
I don't think "a chatbot isn't AI and hasn't been since the 1960s when they were invented, whether you call it a doctor or a Ukrainian kid doesn't make any difference" counts as shifting the goalposts.
Furthermore, reproducible results are an important part of science. Let him release his source code, or explain his algorithm so we can reproduce it. Anything less is not science.
"First they came for the slanderers and i said nothing."
Gives a whole new meaning to, "My computer went down on me..."
Do not look into laser with remaining eye.
It was passed as defined
The Turing Test was not passed, and the only people who claim it was are ignorant reporters looking for an easy story with a catchy headline
Indeed. There's a lot of misinformation out there about what Turing originally specified. The test is NOT simply "Can a computer have a reasonable conversation with an unsuspecting human so that the human will not figure out that the computer is not human?" By that standard, ELIZA passed the Turing test many decades ago.
The test also doesn't have a some sort of magical "fool 30%" threshold -- Turing simply speculated that by the year 2000, AI would have progressed enough that it could fool 30% of "interrogators" (more on that term below). The 30% is NOT a threshold for passing the test -- it was just a statement by Turing about how often AI would pass the test by the year 2000.
So what was the test?
The test involves three entities: an "interrogator," a computer, and a normal human responder. The interrogator is assumed to be well-educated and familiar with the nature of the test. The interrogator has five minutes to question both the computer and the normal human in order to determine which is the actual human. The interrogator is assumed to bring an intelligent skepticism to the test -- the standard is not just trying to have a normal conversation, but instead the interrogator would actively probe the intelligence of the AI and the human, designing queries which would find even small flaws or inconsistencies that would suggest the lack of complex cognitive understanding.
Turing's article actually gives an example of the type of dialogue the interrogator should try -- it involves a relatively high-level debate about a Shakespearean sonnet. The interrogator questions the AI about the meaning of the sonnet and tries to identify whether the AI can evaluate the interrogator's suggestions on substituting new words or phrases into the poem. The AI is supposed to detect various types of errors requiring considerable fluency in English and creativity -- like recognizing that a suggested change in the poem wouldn't fit the meter, or ir wouldn't be idiomatic English, or the meaning would make an inappropriate metaphor in the context of the poem.
THAT'S the sort of "intelligence" Turing was envisioning. The "interrogator" would have these complex discussions with both the AI and the human, and then render a verdict.
Now, compare that to the situation in TFS where the claim is that the Turing test was "passed" by a chatbot fooling people. That's crap. The chatbot in question, as parent noted, was not even fluent in the language of the interrogator, it was deliberately evasive and nonresponsive (instead of Turing's example of AI's and humans having willing debates with the interrogator), there was no human to compare the chatbot to, the interrogators were apparently not asking probing questions to determine the nature of the "intelligence" (and it's not even clear whether the interrogators knew what their role was, the nature of the test, whether they might be chatting with AI, etc.).
Thus, Turing's test -- as originally described -- was nowhere close to "passed." Today's chatbots can't even carry on a normal small-talk discussion for 30 seconds with a probing interrogator without sounding stupid, evasive, non-responsive, mentally ill, and/or making incredibly ridiculous errors in common idiomatic English.
In contrast, Turing was predicting that interrogators would have to be debating artistic substitutions of idiomatic and metaphorical English usage in Shakespeare's sonnets to differentiate a computer from a real (presumably quite intelligent) human by the year 2000. In effect, Turing seemed to assume that he would talk to the AI in the way he might debate things with a rather intelligent peer or colleague.
Turing was wrong about his predictions. But that doesn't mean his test is invalid -- to the contrary, his standard was so ridiculously high that we are nowhere close to having AI that could pass it.
I think Watson would be able to give it's real age by finding the information rather than recalling it, although it might get confused by progressive versions. AI can also produce a picture of a generic rabbit, or cat as the case may be.
The thing that Watson (and AI in general) has difficulty with is imagination, it has no experience of the real world so if you asked it something like what would happen if you ran down the street with a bucket of water, it would be stuck. Humans who have never run with a bucket of water will automatically visualise the situation and give the right answer, just as everyone who read the question have just done so in their mind. OTOH a graphics engine can easily show you what would happen to the bucket of water because it does have a limited knowledge of the physical world.
This is the problem with putting AI in a box labeled "Turing test", it (arrogantly) assumes that human conversation is the only definition of intelligence. I'm pretty sure Turing himself would vigorously dispute that assumption if he were alive today.
And did you exchange a walk on part in the war for a lead role in a cage? - Pink Floyd.
One of my friends is a philosophy post-doc and he told me many times that in philosophy the gold standard for intelligence is intelligent behaviour. Of course he has some footnotes to add, notably that intelligent things can appear to be bricks if you cut off all their actuators, but to say that this particular variant of ‘behaviourism’ as you call it is discredited is disingenuous. In particular, if one could hypothetically replace someone's brain with a computer and not know the difference then the computer must necessarily be intelligent, insofar as humans are intelligent. Being a philosopher he has opinions about that too.
It also isn't true that the Lovelace test is more rigorous. To pass it you must produce something truly original but presumably non-random. I can only say good luck getting any human to pass this test. In practice this means of course that the bar must be lowered to some measure of non-rigorous relative originality. The weird use of the word ‘program’ doesn't make the fact that we're could actually be talking about a poem any more rigorous either. Then there is the strange notion that the machine's designers must not be able to explain why it works the way it does. Quite apart from the fact that a lot of software has multiple shades of this already as it is, the designers of course wouldn't be able to explain even the most trivial action their machine took. Of course you have to broaden this from ‘the designers’ to ‘anyone’ but then you get back into the problem that if someone ever figures out how the brain works, yours truly will no longer be intelligent according to our dear Lovelace test.
And as a matter of practical importance, there already exists a lot of machine generated art, some of it quite beautiful. The programmers can explain the algorithms of course, but often not how it got the final result just right. These would appear to pass the Lovelace test (you can disagree if you want but that just shows that the Lovelace test isn't rigorous), but are not intelligent to the best of our knowledge. The Turing test, as envisioned by Turing (rather than the garbage that has been in the news lately) is a much more reliable measure of intelligence. Could there be a better test? Probably. Maybe something is only intelligent if it can get a philosophy degree. Too bad that this test would disqualify almost all humans, though.
No, he's not describing behaviourism. He's saying this:
If it behaves intelligently, then it is intelligent.
That's a reasonable statement to make, and if you're disagreeing with that statement, you need to say why. Converting it to a strawman and then making a bald claim that the strawman is "discredited" is a cheap rhetorical trick. And then you go on to talk about free will, which has no direct relationship with intelligence anyway. OK, I get it, you want to turn the conversation around to being about free will, because that's your ax, but telling someone their perfectly reasonable statements are "simply wrong" is a shitty way to do it. OP's point, which you're deliberately missing, is that whatever intelligence is, it is not an observer-relative thing which demands that the observer be unaware of the mechanism. If you want to engage in debate with him, try addressing that specific point, rather than a bunch of points he never made about a subject he's not discussing. And if you want to talk about free will and about how behaviourism is "discredited" maybe you could at least make a couple of points in favour of that argument, for those of us who might be interested anyway. Maybe then we can see how your belief relates to what is actually being said.
Anyway, what you're both missing is the practical issue with "The machine's designers must not be able to explain how their original code led to this new program." The machine's designers can lie, or be incapable of coming up with an explanation despite one existing, so this is a completely ill-defined criterion - which is what we're trying to get away from.
Turing was wrong about his predictions. But that doesn't mean his test is invalid
Imho it is.
Suppose we manage to create a strong AI. It's fully conscious, fully aware, but for some quirk we cannot understand, it's 100% honest.
Such an AI would never pass the Turing test, because it would never try to pass off as human, and any intelligent human could ask it questions that only a machine could answer in limited time.
Turing was wrong about his predictions. But that doesn't mean his test is invalid
Imho it is.
Suppose we manage to create a strong AI. It's fully conscious, fully aware, but for some quirk we cannot understand, it's 100% honest.
Such an AI would never pass the Turing test, because it would never try to pass off as human
That sounds like a legit point at first, but think about it for a sec. Programming a computer to lie and be evasive about its nature is easy, and many chatbots can already do that. Asking a strong AI "are you a computer?" or "what did you have for breakfast?" would not be useful for evaluating intelligence. Getting the AI to debate an intellectual topic, on the other hand, will be less likely to require deception but would be a better measure of intelligence. That's another fundamental point people miss: The point of the Turing test was to imitate human INTELLIGENCE, NOT to pretend to be a physical human.
A knowledgeable interrogator trying to evaluate intelligence would thus likely be more interested in asking intellectual questions, rather than queries just designed to test whether the computer can make up some nonsense about itself.