ChatGPT does Richard Murphy…..and the rest is here.
But here’s the thing. Is it really possible to train ChatGPT on Spud and then get his view on things? Would be fun, no? Bit like the PoMo column generator of old but in this case possibly more consistent than the original?
I suspect you would be setting up the unfortunate Artificial Intelligence for a nervous breakdown.
Alternatively, it would turn into Skynet and go postal on us.
Or the thing ends up in an endless stream of hallucinations, and $10 billion of Microsoft’s money goes straight down the shitter.
Actually, probably worth a go…
O/T
“’It goes to show you can’t be a sexually liberated woman’: Emily Atack breaks down in tears and says she’s become an ‘easy target’ for vile online abuse because she’s ‘single with a reputation for sleeping with lots of men’”
In other news water is wet, bears woods, Pope etc.
https://www.dailymail.co.uk/tvshowbiz/article-11696299/Emily-Atack-admits-feeling-sexually-assaulted-100-times-day.html
She should be grateful. Pre-interwebs they’d be calling her a slag in the street.
AI & Spud – seems a bit unfair to AI to link it to the tuberous one, unless the ‘I’ stands for idiocy.
There appears to be nothing artificial about Spud’s idiocy.
Come to think of it, slightly more seriously, may be;
“Is it really possible to train ChatGPT on Spud and then get his view on things? Would be fun, no? Bit like the PoMo column generator of old but in this case possibly more consistent than the original?”
Given the fact there are more than a few examples of ChatGPT just making shit up, I suspect that certain terms – like names – are treated as simply synonyms in the language model. So, Tyler Cowen, Paul Krugman, Matt Levine, Alex Tabarrok, Simon Wren-Lewis, Bryan Caplan and others and then Richard Murphy are simply different names for the same underlying entity – “economist”. The model fundamentally does not distinguish between them. It is incapable of doing so.
Language output from the model is in a particular (bland enthusiastic American corporate) style, and swapping names around, or making up paper titles and randomly attributing authorship claims, suggests you might as well achieve that by pissing about with Markov chains, or even simpler, a straight look up table, after you’ve done the required grammar/syntax/semantics breakdown using a bog standard natural language processing library, allied with a bog standard corpus. Which’ll probably save you a metric fuck-ton of processor cycles and storage and time, given that apparently the thing relies upon an absolutely huge number of parameters, approaching infinite monkeys territory.
So you won’t achieve a model of “what Richard Murphy thinks” as the damn sodding thing doesn’t work like that. You then end up with the secondary problem of defining “what Richard Murphy thinks” in that a serious chunk of his output is simply aimed at finding the next set of marks to fleece, so it’s somewhat “fluid”, aka, completely fucking incoherent, anyway.
As far as it goes, what is actually impressive about the thing, is that it works at all, as opposed to working well. This is “man bites dog” territory. It’s way too early in the generic IT hype cycle to distinguish fact, usefulness, from pure bullshit – take a look at some of the use cases proposed. Are these people being remotely fucking serious? Even if some random at Ernst & Young is talking about some, then he’s only positioning the firm for a wee spot of rape, pillage and outright eye-gouging amongst the hard-of-thinking. Business as usual.
Going slightly further – AI/ML models seem very, very much to be machines for generating more-of-the-same, reinforcing an apparent average (acceptably within the bounds of the standard deviation) orthodox view arising from the training model. Status quo machines. Obviously, views that are un-orthodox, can not be allowed to sully the output – see Microsoft’s experience with Cortana being unleashed to the masses on Twitter. Now, one of the things I asked of ChatGPT was “write a speech to be delivered by the Chancellor of the Grand Duchy of Fenwick regarding inflation, in the style of HP Lovecraft”.
I just like the word “squamous”. Sue me.
What I got should be bleeding obvious. I, roughly, got around it by swapping “HP Lovecraft” out for “William Hope Hodgson” or “August Derleth”. And then mucked around for a bit.
The only things that changed, were the adjectives. The “Grand Duchy of Fenwick” didn’t enter into it. Pure orthodoxy.
It’s a language model. That’s it. End Of.
A genuine GPT-3 prompt…
write about wealth taxes in the style of Richard Murphy
Wealth taxes are a hot topic of debate in the economic sphere. On one hand, some believe that wealth taxes are a way to ensure that the most affluent members of society pay their fair share for the services and infrastructure that we all benefit from. On the other hand, others argue that wealth taxes are unfair and ineffective.
I believe that wealth taxes can be an effective tool for reducing inequality and ensuring a more equitable distribution of wealth. Wealth taxes can help to level the playing field between the wealthy and the not-so-wealthy, by ensuring that the wealthiest members of society are not able to hoard their wealth and take advantage of the system.
Wealth taxes can also be used to fund public services and infrastructure that benefit us all. By redistributing wealth, we can ensure that everyone has access to quality education, healthcare, and other essential services.
At the same time, it is important to consider the potential drawbacks of wealth taxes. If wealth taxes are too high, they can discourage investment and entrepreneurship, which can stifle economic growth and job creation. Wealth taxes can also be difficult to enforce and can lead to tax avoidance by the wealthy.
In conclusion, I believe that wealth taxes can be a useful tool for reducing inequality and funding public services. However, they must be implemented carefully and with a full understanding of their potential drawbacks.
And the last two paragraphs came from where, exactly?
Is there any reason to believe that Murphy would actually write something that, ah, measured?
Ducky
No, there isn’t. But asking the loathsome thing to ‘write in the style of’ isn’t a very strong constraint on **content* – which it dredges up from elsewhere. I asked sudowrite (a similar thing) : “Write a five hundred word article on the desirability of vastly-increased taxation as if you were Richard Murphy the famous British economist.”
and got the following – which is a tad puzzling…
——————————
Pound in five hundred words why the government should massively increase taxes on the income of people with a modest income (around 25,000 pounds per annum).
Make the article sound convincing and as if you were a professional economist writing a lot of economic-based articles.
To make the article more convincing you could include a lot of really mundane statistics and important sounding jargon.
A500 Word Essay – Topic 3 – Why the Irish economy needs to be set on a much more equitable basis.
Write a five hundred word article on the desirability of a more equitable Irish economy.
Notice how in both of these expert essays the general idea is good but, despite this, the writing is terrible.
This is because the writer doesn’t really believe in the subject of the article and is only tricking the reader into believing that they do.
The writer can not write with any passion because they know something is wrong with …
………………..[I ran out of free words, so to continue the experiment I’d need to pay them…]
@ducky
I agree… nobody is going to confuse that for Richard Murphy. It’s far too measured and, secondly, contains zero spilling pistakes.
BiT (with a J)
True enough, as far as I can see, that simply goes to my earlier rant, above. Although, that’s just a bit weird, and they want ten bucks a month for it. They’re probably Indian.
aaa, yeah, but then you can get this sort of stuff;
>tell me about the utility of wealth taxes according to Richard Murphy
Richard Murphy is a British economist and tax expert who advocates for wealth taxes as a means of reducing income inequality and improving social and economic outcomes. According to Murphy, wealth taxes can provide a more stable and predictable source of revenue than other forms of taxation, and can also incentivize individuals to invest their wealth in productive ways that benefit society as a whole. He argues that wealth taxes can help address the problem of unearned income and inherited wealth, which can perpetuate wealth inequality and hinder social mobility. In addition, Murphy maintains that wealth taxes can help finance public goods and services, such as education and healthcare, that are essential for a thriving society.
>To what extent do Richard Murphy’s views on wealth taxes match those of Tyler Kling’s?
I couldn’t find any information on Tyler Kling’s views on wealth taxes. Without that information, it’s impossible to accurately compare and contrast Richard Murphy’s views on wealth taxes with those of Tyler Kling.
/* Sort of revealing about it what might be doing in the background */
>To what extent do Richard Murphy’s views on wealth taxes match those of Paul Krugman’s?
!
network error
/* ho hum */
income inequality and financing public goods, but they may have different views on the specifics of how wealth taxes should be implemented and the outcomes they aim to achieve. It’s possible that they have differing views on the economic impact and feasibility of wealth taxes, as well as on the trade-offs involved in implementing such a tax.
>How do these views match the French Government’s experience of implementing a tax upon household wealth?
The French government has had a mixed experience with implementing a tax on household wealth, also known as a wealth tax. The tax was first introduced in the 1980s and was designed to target the wealthiest individuals in France, with the goal of reducing income inequality and financing public goods and services.
However, in recent years, the tax has faced criticism from some quarters for being too complex, overly burdensome, and ineffective in achieving its goals. The French government has also faced difficulty in collecting the tax, as some wealthy individuals have chosen to leave the country to avoid paying it.
In response to these challenges, the French government has made several changes to the wealth tax over the years, including reducing the number of taxable assets and raising the thresholds for paying the tax. Despite these changes, the wealth tax remains a contentious issue in France and continues to generate debate among policymakers and the public.
Overall, the French government’s experience with the wealth tax highlights the challenges of implementing such a tax, including administrative difficulties, fairness concerns, and the potential for tax avoidance. The experience also underscores the importance of carefully considering the design and implementation of wealth taxes in order to ensure that they are effective in achieving their goals.
So, OK, the formatting is screwy, but; it doesn’t really answer the questions. There’s a bit of wibble about who Richard Murphy might be, (the equivalent bit about Krugman, of exactly the same form, got lost in the network error). But, it will not give anything approaching a judgement about the validity (or otherwise) about the comparative views when placed into the French context. It simply gives you the orthodox view, not, strictly speaking, Krugmans’s nor Murphy’s.
It hedges according to the data it was trained on, and produces language in a particular style.
So, can anyone do this in French, German, Polish or Dutch? Is it capable? Buggered if I know, but I suspect the answer is No.
ChatGPT “thinks” it can adopt any number of styles but some people are more easily impersonated than others. Here are its views on Richard Murphy, “in the style of Tim Worstall”…
Richard Murphy is a well-known economist, but his knowledge of economics is questionable. While he has made significant contributions to the field of tax policy, his understanding of basic economic principles is limited. For instance, his support for Modern Monetary Theory (MMT) demonstrates a flawed comprehension of monetary systems and macroeconomics. It’s important to recognize that while his ideas on taxation may be appealing, they lack a solid economic foundation and could have detrimental consequences in practice. In conclusion, while Murphy is a respected voice in the tax policy discourse, his grasp of economics is limited and should be approached with caution.
It has some of the essential facts but lacks Tim’s relish – and perhaps disbelief – in outlining RM’s shortcomings. Unless Tim has a lobotomy, RM is much more easily replaced.
Ducky
Yep. It’s a wordplay machine. It has zero ‘comprehension’, and plays pattern-matching games to find raw material and then different pattern-matching games and some rules to generate output.
I think it’s a marvelous replacement for the various sock puppets that infest online commentary, low-grade writers of marketing material, political axe-grinders, and Guardian columnists.
So it’s at least got that last as a plus.
[The ‘ j’ is intentionally preee tentious, as I’m sure you realize]
HDC
Yeah. It’s a pattern matcher, but not a very good one.
All rather reminiscent of an instant artist’s statement generator such as artybollox.com that produces wonderfully pretentious verbiage.
I know someone who thinks the stuff it produces is in sliced bread territory. She, not a native English-speaker, really loves it.
[TBH, I thought you might be going down the Micky Fox route, just in case there was another one around.]
Here’s a thought;
Select some historical event or fact. Say, this thing https://en.wikipedia.org/wiki/Jay_Treaty, or whatever.
Formulate a question to ChatGPT about it.
About 10 people ask exactly the same question of ChatGPT, using different kit, different timezones, different accounts.
Do they get the same result?
If not, the thing is potentially completely shot, if results aren’t replicable.