To Survive the New Era of the Spam Bot, Watch the History of Cryptos

[ad_1]

Here’s David Z. Morris, filling in for Michael Casey to talk about so-called artificial intelligence, the threats it poses to the future, and how crypto could help mitigate them.

As Michael would surely agree, there are no real days off in crypto. I remembered when I recently spent a long weekend at the fantastic Readercon fiction convention. Inevitably, I missed some important crypto stories, but I also got a close look at another looming novelty: the existential threat that large Automated Language Models (LLMs) like GPT3 pose to the entire Internet.

You’re reading Money Reimagined, a weekly look at the technological, economic and social events and trends that are redefining our relationship with money and transforming the global financial system. Subscribe to receive the full newsletter here.

This may sound hyperbolic. But at Readercon, I met Neil Clarke, founder and editor of the leading science fiction magazine Clarkesworld, who, along with other fiction publications, has become a canary in the coal mine of AI run amok. The rise of ChatGPT has flooded these reviews with a flood of fake GPT-generated story submissions, a plague so severe that Clarkesworld was forced to temporarily suspend submissions in February, threatening the jobs and livelihoods of real authors.

I call it spam, Clarke said, because that’s what it is. Sometimes I even refuse to call it artificial intelligence. You can’t humanize these things. It’s not like science fiction movies where it’s conscious. It’s a statistic [language] model.

The mention of spam should raise the air of long-time cryptocurrency watchers: the same problem is at the very root of Bitcoin.

Between 1998 and 2002, computer scientist Adam Back developed the concept of Hashcash, primarily intended to combat email spam by requiring a small payment to send one. Back and his ideas have become fundamental to the development of Bitcoin, and he is now CEO of crypto developer Blockstream.

Two decades later, with hordes of barbarian robots about to flood human communication systems, it may be time to revisit the concept of Hashcash.

ChatGPT was released in late November, Clarke says, and we immediately started seeing submissions using it. The first to adopt it are those who were already submitting plagiarized works. It was easily adopted by people trying to make a quick buck from the work of others.

As they dealt with the spam problem, Clarke says he and his team quickly realized the attack was coordinated. YouTube and TikTok channels focused on get-rich-quick schemes promised viewers they could earn thousands of dollars by submitting GPT-generated stories to fictional magazines like Clarkesworld. Clarkesworld pays a few hundred dollars per story, depending on the length not much more than beer money in some parts of the world, but hugely significant in others.

These fraudulent promises of online scammers seem to have spread rapidly. Clarke says he received 54 AI-generated submissions in December. In January, he had 117 fake stories. In February, the number hit 514 before Clarke closed submissions at noon on February 20.

And that morning alone, he says, we had 50.

Clarkesworld has a small team, normally reviewing around 1,100 submissions per month. Thus, the accelerating flow of waste threatened to overwhelm them, and the solutions were not obvious.

We have an open submission process, specifically designed to welcome new writers and new voices, says Clarke. So we might close submissions from some places [to fight spam], but we also have legitimate authors from these countries. And we were told things like, Paying for this story will cover my bills for a month.

Authors like that are buried. AI submissions hurt new authors and authors who might not be from well-connected communities. This is a clear way that auto-generated content threatens to make the internet worse for human beings, especially those on the fringes.

If you go back 15 or 20 years when we were taking paper submissions, Clarke says, the cost of postage alone was enough to significantly reduce submissions from outside the US, Canada and the UK. And as soon as you have digital submissions, we had this flood of international submissions. This led to a huge diversification of the world of fiction, a creative renaissance that is now threatened by the rise of LLMs.

Clarke is also a coder, which gave her some useful tools to tackle the spam challenge. It began associating more metadata with submissions, such as whether they came from a VPN and users’ session length. These and other criteria are now used as part of a point system that places stories more likely to be fake further down a review queue. This helps real authors get read first, but also ensures that every submission is ultimately reviewed.

Finally, if a story is determined to be generated by LLM, the author is permanently banned from the system.

These measures have helped Clarkesworld reopen submissions, for now, but a continued increase in the volume of spam they face would mean the fix is ​​only temporary.

An important aspect of Clarkes’ experience is that the actual quality of robotic submissions has been extremely low. They are almost instantly recognizable to a human reader and have no real chance of being published.

ChatGPT3 was writing at a lower level than the worst human writers, says Clarke, who after two decades as an editor knows exactly what the worst looks like. GPT4 comes close to the worst human writers, but even that is rare.

The common thing is that they have perfect grammar, they have perfect spelling, Clarke continues. But the stories themselves don’t make much sense. They skip over important things. They’ll start with a basic premise, like ecological collapse, and introduce scientists who can solve the problem, and then all of a sudden they’ve solved the problem. The middle of the story is missing, and the book with stereotypical openings and closings, very poorly done.

This sounds a lot like Ted Chiang’s recent characterization of ChatGPT’s release as a fuzzy JPEG of the Internet. This overt shit happily debunks much of the hype around LLMs. But it also makes the image of talented (and hugely underpaid) editors having to sift through the dross all the more depressing.

Another option to reduce spam submissions is a submission fee. Clarke says he has no desire or plan to implement a fee, thanks to overriding ethical and creative concerns. In particular, submission fees could limit access, which the science fiction writing community strongly opposes.

But beyond that, technical shortcomings in the current global payments infrastructure also make it impossible to charge anti-spam fees, even if Clarke wanted to.

For example, Clarke says that to charge submission fees, he should be able to reimburse them, for example, to writers whose stories have been accepted, or simply not AI-generated. Ideal spam blocking fees would also be quite low, certainly well below the $25 or $30 postage that kept authors away from developing countries in the pre-Internet era.

But there’s no way to do that with current technology.

Tell me a credit card company where I can refund almost anything. I would lose count, said Clarke. As any good crypto brother knows, credit cards don’t work well with small payments. But that’s not even the biggest problem.

There are also issues with trying to accept payments in different parts of the world, Clarke continues. There are a number of African countries that credit card companies will not work with. This would therefore eliminate the perpetrators. I’ve also had people suggest ID services, but those also have nationwide holes. We need something that works for everyone.

If you’re reading this, you already know where we stand: at least in principle, cryptocurrency and related systems could help alleviate Clarkesworlds fake submission problem.

Requiring a small payment for all submissions would reduce low-quality submissions, ease the workload on publishers, and compensate them for the spam that has arrived. Because payments could be cheaply, quickly, and easily returned to real writers, the cost to real human writers would be marginal. And because these systems are not limited by national borders, no real writer would be squeezed out by bot-regurgitators.

Although this would require considerable elaboration, some versions of the same system may one day serve parallel purposes in less specialized environments. One can imagine a Steem-like staking incentive system being used to punish automated posting on forums or social media, for example. More elaborate decentralized identity systems, such as SpruceID, are more challenging and, for now, more nascent, but could have even deeper potential.

To be clear, none of this should be necessary. LLMs are quickly proving to be little more than parlor tricks, the real usefulness of which is likely limited, at least in the short term, to abbreviated customer service and clickbait entertainment. (Take for example CNET’s disastrous experience with using GPT to write news articles).

Rather, the greatest impacts of technology are seen in the spread of fourth-order gibberish that wastes the time and intelligence of all humans involved. But if this is what the god-princes of Silicon Valley see as the next frontier of venture capitalist riches, then this is the world to live in. At the very least, the crypto offers some hope of fighting back.

[ad_2]