Become a fan of Slashdot on Facebook

 



Forgot your password?
typodupeerror

Comment Re:This is not an AI failure (Score 1) 148

That folks using these pattern matching and regurgitation systems don't realize "Do not delete the code" and "I will not delete the code" have null meaning to a system, other than the value of weights based on the words and letters.

There is no "I" to understand what these strings of letters and words actually mean.

It just regurgitates common pattern completion it's ingested after quantitizing "Do not delete the code". It sees "I will not delete the code" as the most common words/order to follower the inputed words.

AAAAAAAUGH!!!

From the time of throwing bones and looking at patterns in the stars stupid human-apes have wanted answers from something other them themselves.

Exactly, and well said!

At least with traditional programs based on running logic rules, the lack of an "I" (which can check with other felt experiences, like, did I just hammer my own thumb) isn't a problem as much, if the rules are well defined enough, and we're not getting tripped up by logic flaws, like unanticipated coercion of values etc.

Crucially, it is the programmer who is the "I" performing the hard work of representing the real world using symbols and rules, of creating a working logical model representation.

But with AI, the "I" programmer is gone and instead we're relying on it training itself on a vast dataset — and assuming that the resulting working model isn't just a machine but somehow "understands" what it is doing — this was never the case with traditional programs, and so far isn't the case with LLMs either? But because the method is training, rather than a human doing the work of figuring out the logical representation model, we actually end up with less reliable models?

Hence the need for humans to find ways to retrain and bias the resulting correlations, basically beat it with a stick when it gets things wrong, and we rely on people finding all of those problems, and hope the model remains stable enough to not have that forcing mess up some other correlations.

If I understand correctly.

Comment Re:This is not an AI failure (Score 2) 148

The hyperbole, around general intelligence models replacing humans, is a massive, massive distraction from their real value.

Their real value is to apply these deep learning models in a very domain specific way, get them to solve very specific problems; analyzing voice, images, picking out patterns in data, etc.

We seem to have gone down a very weird deluded story which is boosted by the delusion that, because these models can sound and talk like a human, that they can suddenly replace a thinking human in all sorts of jobs and areas -- even though a human being is a product of three million years of evolution, and even each job requires years of experience and skill to understand. Many jobs are networks of relationships between people and contexts -- things the people have to learn as they enter the organisation and find their feet. (And why outsourcing or even reverse outsourcing is so often problematic.)

It's a fantastically ironic situation, that people are betting huge sums of money on intelligence, out of their own ignorance about what intelligence is.

What these general LLMs seem to be doing very well, is to do what people often do, which is to hear and repeat the kinds of things which are often said and which are therefore taken to be kinds of truisms and general guides to what the assume and believe about problems and issues.

If you're surrounded by a culture that believes for example that Russia is our eternal enemy, then that's pretty much what you're going to pick up from people around you, from the news, from the media, and it's what will hence also get fed back to you by the LLM. None of that is original thinking.

Comment Re:This is why you have offsite backup (Score 1) 122

For a small company with some on-site stuff, there's tons of options.

First make sure the backup server can access live servers and desktops and not the other way round. That stops an infected server being able to encrypt its own existing backups.

Next, use a file system that does checksums like ZFS. Have a separate isolated machine scrub each drive periodically to ensure that they still work. This just means having someone do it and having a clipboard with a checklist like how cafés have rotas for cleaning the toilet.

Then, put drives in a on-site fireproof safe which gives you 40 minutes in the event of a fire.

A small company may have a couple of other locations which they can just carry disks to as well. If that is too much of a hassle, then use an entirely separate cloud storage area to upload snapshots.

There's going to want to be a temptation to automate everything and then forget about it, but an advantage of having to have all these manual steps is that it means someone has to take care of doing it which means it's being done consciously.

Comment Re:Colonialist looting invasionist Britain is the (Score 1) 76

I cannot comment on the content of your post because I simply don't know the history of India,

and your comment does remind me of a book by Steven Pinker called The Better Angels of Our Nature where he details, in gruesome descriptions, the fact that humanity's past has been filled with barbarities.

We have gradually on everage become more peaceful and more empathetic through our evolution.

Humanity's history is full of horrors, and these happened pretty much everywhere, as far as I understand it.

What's extremely confusing today, I think, is that not everywhere has changed to the same amount of empathy at the same time. Just like in any population or country in the planet, a certain percentage of people are going to be psychopaths.

Where a lot of, perhaps, let's call them liberal views, get it a bit wrong, is imagining that absolutely everyone in the world holds the exact same liberal values and attitudes as the best liberal people of the West do.

Actually, it's a vast tapestry, and some areas are more liberal, and other areas are less liberal. And some people in amongst groups are more liberal, and some are less liberal. Some parts of the world are more violent, in their systems, their attitudes, beliefs, etc. Some do violence in different ways -- USA and Russia have 13,000 nuclear warheads pointed at each other -- that's violence also.

And this is a challenge for globalization, but I think it just means that globalization needs to happen -- to get all its benefits -- gradually, perhaps slowly in a moderated way, rather than massive changes which are perhaps a little bit counterproductive.

There is a common humanity but we also have a common barbarity in our collective history and -- as the saying goes -- the future has already been invented but it isn't evenly distributed yet -- so have to include that in a way that works, not in a way that's chaotic and catastrophic. It just needs doing with moderation.

But this also means being moderate in how we go about making generalizations about groups. That becomes labeling and othering, and that, if done in a blind way, doesn't really help anyone.

It's better to just look realistically at peoples and situations and context. As they are now. As you find them right now. Try to see everything fresh -- we can all still make judgments, we just have to be careful that they are good judgments rather than blind judgments.

Every judgement is only a piece of the picture, a partial truth.

For example, it seems to be in the interests of the world's superpowers to promote and back the most extreme and violent groups amongst developing countries as a way to keep those countries destabilised, fragile, and easier to control and influence.

If the world's superpowers had been actively backing and promoting the most reasonable, modern, forward-looking groups amongst the developing countries, I think the picture today would be quite different.

And that doesn't negate the barbarities of the past from across the world, be it the Roman Empire, the British Empire, various Chinese dynasties, etc.

Comment Re:in other words (Score 4, Interesting) 181

I agree that humans mimic LLMs with respect to probability judgements. Marketers know that if you see a "fact" written in a few different articles or sources, you come to assume it's true, for example. We rely on what our culture feeds us and we internalise it as beliefs.

But the other words you mentioned are actually very difficult and deep questions which smart people throughout the ages have wrestled with and we still don't know the answers today. Sentience/the ability to have an experience is the most obvious and direct reality we each have, yet nobody knows how that works.

Yes our minds can remember things we've heard and repeat them like a photocopier or an LLM, but we don't know what is experiencing the whole show.

Slashdot Top Deals

One man's "magic" is another man's engineering. "Supernatural" is a null word. -- Robert Heinlein

Working...