nmaley

Microsoft's Power Platform is a low code framework which works well, and generates a massive amount of revenue, as does Salesforce and some others. I recently designed and implemented a complex 500 user child protection application with PP that has been live for a year now. It was highly successful, and the time and cost taken to deliver it was far less than the cost of a hand written solution. That said, there is still quite a lot of custom code required for most enterprise level solutions even with the most mature low code platforms. Low code is not a panacea, and the same issues of how to represent requirements and design arise in the low code world as in the high code world. Low code platforms will continue to mature and improve. Maybe AI will catch them one day, but I'd be surprised if that happens anytime soon.

As an exercise, I spent hours and hours and hours re-writing a chapter of my PhD thesis strictly according to Sword's principles. The eventual result was shorter, clearer and more readable. But it was hard work, and I'm not sure the effort justifies the reward for academic writers. For example, this post would score poorly if marked according to Sword's principles. Better to aim just for clear, short sentences.

It's a really deep problem, which bothers a lot of philosophers and cognitive scientists. I don't pretend to know the answer, but here are some speculations.

We can speculate that consciousness is some kind of a representational process, using non-verbal, non symbolic representations. We can suppose that what is being represented is in part the state of my own body, and my current interactions with the external world. So this is not a static type of representation, it is a continuously changing representation, which is representing a real time 'dance' between my body and the world. So, returning to your (excellent) analogy of the field of rocks, the field of rocks would not be conscious unless and until it becomes an emergent representation of itself and its boundary with the rest of the world, all happening automatically and in real time. The field of rocks has no sensation of the sun's warmth because it has no representation of warmth, and no way of moving or changing it's own body as a response to that representation of warmth, and no attentional mechanism to prioritise how it responds to representations of warmth, hunger, thirst etc.

Once you start to think of consciousness as a representational process you start to see how consciousness might become an emergent phenomenon inside a machine 'designed' for creating representations. A machine like a living animal.

Of course, that puts you in the position of having to explain what is a non-symbolic, or non verbal, type of representation. But that's doable, I think.

The relationship between language and the world is this: utterances both signify and depict objects and events in the real world. So, if I say "I saw Alec Baldwin at the bastketball game last night", then that depicts an event in the real world. And, if and only if the statement is true, an event similar to that depicted was part of the causal history of the utterance itself. The causal history of the utterance determines the significance of the utterance, just as the causal history of a footprint determines its' signficance. To understand a sentence is to understand what it depicts in the real world, and what it actually signifies in the real world. The ability to tell true from false is the ability to detect a disconnect between what is depicted and what is actually signified. That is what LeCun implicity means by language understanding.

So, what does a sentence produced by GPT3 signify, and what does it depict? What it signifies (ie causal history) is that this sequence of words is what human writers would most likely use when producing an utterance containing whatever trigger words the model has been fed. In other words, it's a statistical modlel. What it depicts is whatever the mapping rules for that language tell us it depicts. Since human beings usually tell the truth, a statistical model will usually produce true statements. It will also seem to have the ability to tell true from false, in many cases. But because GPT3 has no model for the significance of its sentences, it cannot be said to have any understanding of language, in the sense humans have it. LeCun's point about flying an aeroplane to the moon is essentially correct.

The theory needs a lot of development, so it would be hard to say whether it could or could not comply with Bells theorem. But since it is Turing complete presumably there is some variant of the theory which complies with Bell's theorem.

I find it interesting that under this theory regions of space and properties of objects are emergent phenomena. I suspect that under some variants of the theory, properties of objects could act as if they are both random and non local.

So true! I built a solution for a legal services outfit a few years ago. First bug they see, they want to sue you. Forget about working with the vendor to refine the solution. They just want to sue you. Never want to work with lawyers again.

Has anyone observed that the problem of object classification is really just a component of the prediction problem? That the old lady doing donuts in the wheelchair is identifiable as such because her appearance changes in certain (predictable) ways from moment to moment? In other words prediction error minimisation is the basis of the classification, not something made possible by object classification.

I'm sorry, but this didn't cut it for me. It suggested mostly inappropriate substitutions. It is using a simplistic model, when there are already much better ones available. If you need to inject vitality into your prose, try this instead: http://writersdiet.com/test.php