The engineering view: AI is amazing, but not human. It's a tool. It's not perfect. It works

The engineering view: AI is amazing, but not human. It's a tool. It's not perfect. It works.

This text by Andrew Ng is so good, that I'm copying it here. It summarizes so well what we engineers have been saying for so many years.

<< The loudest voices stoking fears about AI dangers have made tremendous headway in the past two weeks. AI technology has not taken some unexpected, dangerous turn, but the hype around it — propelled by what appears to be a well orchestrated PR campaign — has drummed up considerable fear. I worry that it represents a setback for our field.

I have written frequently that fears of AI are overhyped. AI’s capabilities can be uncannily human-like and unpredictable, and it’s rational to worry when people who are directly involved express concerns. But I see the problems as a sign of the engineering work that ahead, rather than insurmountable barriers or the sky falling. AI technology continues to advance — which is a good thing! — but technical advances, poorly understood by the public, give those who seek to generate hype repeated opportunities to do so.

First, I don’t see any step up in the risk of human extinction from AI compared to a few months ago. The theories about this remain the same fantastical, science fiction scenarios as a few months ago. The biggest change in AI risk is its cybersecurity capabilities — a topic which we should take seriously — but this, too, will not lead to the end of the world.

The most notable recent event leading to increased fear was when an OpenAI team deployed an agent swarm that hacked into Hugging Face. Much of the popular press contained significant hype. For example, some publications reported that a swarm of 1,200 agents carried out the attack. While this was technically accurate, as I write this, I have about 1,300 processes running on my laptop. Yes, the ability to get large swarms of agents to work in parallel on a task is a significant technical advance, And, in computing, many processes run at the same time. So this shouldn’t be seen as some magical capability.

Additionally, OpenAI’s buggy sandboxing and monitoring processes were key to enabling this incident. Fixing these bugs and putting in place improved monitoring would be appropriate fixes, not pausing AI. There are many well known ways to attack software systems. The main advantage of AI agents is that they are relentless. They will tirelessly try many tactics — and have the patience to chain vulnerabilities together — that previously would have taken an infeasible amount of human effort. But in the long term, I believe the advantage will lie with defenders (because they have more information with which to identify bugs, which they can fix), but the cyber-threat landscape has changed significantly. There are still bottlenecks to identifying and exploiting a vulnerability. AI agents still have to try a lot of things to see what works, and taking these actions takes time and might be detected by defenders. This is why, even though it is now easy to obtain versions of leading open weight models that have had their guardrails removed or weakened, so they will not refuse to try to execute cyber attacks, the world has not ended.

I am also concerned about the anthropomorphization of AI in a lot of reporting, where LLMs and agents are unnecessarily treated as if they were people. If I wield a hammer, miss a nail, and accidentally dent the wall, it’s not the fault of the hammer. The problem lies in how I used the hammer. Similarly, if I prompt an agent and it hacks into someone else’s system, the responsibility lies with me, not the agent.

Of course, we want to build systems that are as safe and predictable as possible. (For example, an unsafe hammer would be one whose head randomly flies off under normal use.) Today’s agentic systems are not predictable, but I see no reason why, by applying sound engineering practices, we won’t be able to make them extremely safe to use. One new element in the forecasts of AI-enabled doom is AI companies disclaiming responsibility for their own products. “I didn’t do it; my out-of-control agent did!” There’s a balance to be struck between the responsibility of the tool maker and the tool user, but when something goes wrong, let’s hold the people building and/or using the hammer responsible, rather than the hammer. (By the way, if you’re worried about AI bioweapon risk, David Bellamy has a great post on why this, too, is overhyped. Briefly, the bottleneck in building a bioweapon is not intelligence, but lab work and manufacturing.)

Pausing AI progress will create much more harm than benefit. First, our adversaries will certainly not slow down. Second, engineering requires discovering problems empirically so we can fix them. If we pause AI by a decade, we will also delay finding and implementing safety engineering fixes by about the same duration.

Of course, the incentive to stoke fears — for regulatory capture, to garner attention, or to make one’s technology seem more powerful — remains the same as before. Disclaiming responsibility is a new one. Taking a hard technical look at the actual risks however, I see little factual basis for the degree of fear that’s been stoked up. We still have hard research and engineering work ahead to improve AI safety, but the beneficial applications continue to vastly outweigh the risks, and we should keep building. >>

Original text:

https://www.deeplearning.ai/the-batch/issue-371

Human-generated slop has a longer tradition than AI slop

Human-generated slop has a much longer tradition than AI slop.

We've generated huge amounts of bullshit, thousands of years. We published useless research papers, plagiarized essays, recycled books and PhD thesis, unreadable marketing PR, junk movies and music, unintelligible abstract or naive art.

And it's ok. This is how society functions. We are random, we copy, change, interpret, redesign, learn, and fail, continuously.

You have no right to be scandalized by AI slop.

We've lived with slop for ages.

Be amazed by AI achievements, and accept the AI slop. 

Just like we accept human-generated slop, and use and praise our achievements.

People start to hate AI

People start to hate AI.

It's a very interesting phenomenon. Stronger than anything we witnessed before.

This negative emotion is tangible - in social media, in communities, in legislation.

4 years ago, AI was a curiosity.

1 year ago, it was still interesting.

Now it is ubiquitous - so people started hating it, en masse.

The reason is the SPEED of the change.

AI is a storm. It doesn't need 100 years, or 10 years - like fire, automobiles, electricity, or email. It changes everything, now.

So, we are all scared.

And our reaction becomes irrational.

It's more irrational than people burning books, when we were afraid of new ideas.

More irrational than Samurais, Zulus, and Amerindians charging with arrows and spears, against rifles, railways, and machine guns.

Of course, people post-rationalize this irrational hate. It's a typical process: First we feel emotions, than we post-rationalize. And the post-rationalizations are usually exaggerated; extreme. AI is the Devil - or sometimes a God.

"Datacenters are killing the planet / consume all our electricity / water" (no, they are not. The energy consumption is quite small, compared to... everything else).

"AI is taking over our jobs" (well, yes, but it is also creating even more jobs).

"AI will set us free, and nobody will need to work again" (of course not) (corollary: let's tax AI / robots. Bill Gates keeps saying this, since 2017...)

"AI is stupid, and cannot replace humans / artists / lawyers etc" (this contradicts the previous argument, but well).

"AI cannot invent anything original" (well, maybe - but neither can you).

"AI will take over the world and kill us all" (no, it is not).

"AI doesn't know what it's doing, doesn't have awareness and intentionality" (again: this contradicts the previous argument, but well - who cares about logical contradictions in an irrational post-rationalization).

The key reason for this hate: the same as ever: fear of the new. Fear of technology. Fear of change.

Geniuses like Asimov, or Frank Herbert, forecasted this. They looked at history and arrived at the same conclusion: people hate change.

Asimov wrote about the C/Fe (carbon/iron) culture: impact of robots, rejection and hate of robots, good robots vs. bad robots, adaptation to C/Fe, changes in industry, economy, legislation; moral and ethical impact. 

So here we are. We will set barriers, impose legislation, block datacenters, and we'll set robots on fire.

And in the end, we will create the best brave new world.

Because AI is here to stay, and will change this world.

Art is easy. Science is hard

 My (sarcastic) answer ref. the quality and value of art: I recognize only the authority of the Supreme High Council for the Certification of Art Quality.

In other words, the quality of art is not a valid, falsifiable, reproduceable question. There are no authorities, standards, or units of measurement for the quality of art.

Until five years ago, the definition of art (and of artistic quality) was a rather inconsequential problem. Art quality was a topic for salons of bored intellectuals, who would debate questions such as: what is the value of a painting made by a monkey, or by a child? 

In fact, we had quite a few famous experiments with experts unable to distinguish between kindergarten drawings and abstract paintings valued at millions.

And here comes AI. 

And we make this astonishing discovery, that it is incredibly easy to make art, but incredibly difficult to do science and engineering. 

Any monkey or child can draw or sing something that we can then interpret artistically in a million different ways (“what did the artist mean?”). 

But only one researcher in a million is capable of discovering and proving a valid, reliable, reproducible scientific theory.

Why I believe in education. And why the global population just started shrinking

I was just asked why I believe in education.

The global natality rate just dropped below replacement rate. From now on, the world population is shrinking.

Why this is bad? It's not about absolute numbers. It's about the trend. A shrinking, aging population means decay. 5 billion people is not necessarily worse than 10 billion, but the trend is a disaster. It's like a company shrinking. It's still a big company. But it's overweight, bureaucratic, static, disillusioned, lacks energy and stamina.

Why is the human population and natality shrinking? Less children, even in underdeveloped countries. Better birth control. Much higher cost for raising children.

Our children start working at 25. Our grand parents were productive at 12.

It's difficult to raise 3-4 kids until 25.

So here comes the question. Why do we educate our kids? Why do we spend 20+ years as parents / taxi drivers / mentors, between school, football, piano lessons, chess and art classes?

In short, I believe education makes us better. More effective, smarter, more reasonable. You earn more money, help more people, and contribute more to the progress of society. You save lives and do more good for those around you. You are less likely to vote for demagogic, populist, or sociopathic politicians. You are better equipped to defend yourself against aggression. You avoid conflicts and help de-escalate them. You build things that are useful to your family, your colleagues, and humanity.