Ok, allright, AI might be useful after all
As a seasoned sceptic of everything blasted into the ether by Silicon Valley hypemen, I found myself in a bit of a quandary this week.
No doubt, you've seen hundreds of posts touting how this new generation of AI will change the world. Normally, my response would be 'nah this is silly, let's focus on real things.' But this time I've found my head turning, as they say on the reality show Love Island.
For all the hype and predictions, it seems like the technology has matured to a point in which it might actually start being useful for supporting us journalists in some tedious and laborious tasks. It pains my contrarian heart to say this, but I'm rather excited.
During the previous AI hype-cycle circa 2016, a few news outlets –– e.g. Bloomberg ––?started experimenting with the use of AI text generators to cover rote news like stock updates and the weather. It was basically a way to create fluff words around numbers and seemed to do adequately well.
The current generation of Large Language Models (LLMs) are different. While they're being described as 'advanced autocomplete' or more nerdy 'stochastic parrots', the fact remains that the operations these models can perform on text have become quite sophisticated.
Even as the debate has moved on from the performance of these models to a more philosophical question of whether the models 'understand' the text they process and answers they output, it seems somewhat unavoidable that many of us will be using them in the short term.
But will that cost us our jobs? A paper published this week looked into the 'exposure' a range of different jobs had to AI, with unsurprisingly jobs like 'Meat, Poultry and Fish Trimmers' being unaffected and jobs like mine being 100% exposed.
Also unsurprisingly, news media ran with this and predicted doom and gloom with up to 20% of jobs being made superfluous –– which makes for a better headline than 'Workers in some industries might need to work up to 50% less', which would have been a more accurate representation of the paper's findings, and actually sounds pretty great.
Because here's the kicker and also why I'm now more convinced these LLMs will be useful; I think that they're more akin to moving from doing calculations on paper to doing them in spreadsheets. Productivity software, as Ryan Broderick puts it in his excellent Garbage Day newsletter.
When applied properly, productivity software can make more of us, well, more productive. Just like how spreadsheets make it easier for more people to make graphs or whatever, specifically trained LLMs can do the same for things like combing through huge amounts of data or text –– a thing journalists regularly have to do.
It allows us to have the model ingest huge amounts of text, and ask it questions about the text in natural language. LLMs could give the powers trained data journalists wield to regular boring journalists like me who don't know how to use Python. And that makes me tremendously excited, to be honest.
In the past dozen years as a journalist, I've had so many questions that in theory should be answerable because the data is available, but I simply lacked the ability to execute the process of finding an answer.
These questions range from the innocuous ––?e.g. is there way to tell from a picture of the outside of a mandarin if it tastes good – to the more consequential ––?e.g. do Parliament attendance records of political groups show a correlation between attendance and policy.
It would also allow, like I tried out yesterday, to feed a model a piece of legislation, and ask it simple questions about the legislation that would be otherwise hard to tease out.
Naturally, as journalists, even with this productivity software, it remains our task to check facts and verify information, but it can provide a good starting point for asking questions ––?which is another other part of our job.
I think we're not far away from a world in which I (and other curious people with burning questions) will be able to get a somewhat reasonable answer. Although that's highly dependent on how these models will be monetised by the big tech companies producing them, of course, which remains to be seen.
In other news, this week we published a long-ish article on another hype; hydrogen. Wester van Gaal dove into the German hydrogen lobby and how it is affecting policymaking on the EU level . It's a fascinating read, and a good primer for picking up arguments to throw at people who think switching to hydrogen for applications like powering cars or heating homes or even, god forbid, flying planes is a good idea. It's not.
Onwards to the stories you should not have missed this week:
Green Economy
A new report by Corporate Europe Observatory details how German business interests have shaped German and European hydrogen policies.
Health & Society
领英推荐
Two consultations launched this week by the EU to give the public a say on what to do about toxic so-called 'forever chemicals. They could see up to 10,000 chemical substances banned — but the process may take forever.
Rule of Law
The bitter dispute over the so-called 'child protection' law, a red line for the Viktor Orbán government, could also influence the outcome of negotiations on unblocking EU funds for Hungary.
Migration
The Spanish government denies any responsibility over the deaths of some 23 people who attempted to cross from Morocco into its north African Melilla enclave last summer.
EU Political
We're trying out a new format, in which we publish an op-ed and a counter-op-ed. This week, we have the S&D group president responding to EPP chairman Manfred Weber's call for stricter financial screening of NGOs to combat corruption.
Ukraine
The lack of Russian opposition to the Russo-Ukrainian War is puzzling. The war is going nowhere, Russian casualties are staggering, the economy is in trouble, and living standards are declining, and yet polls indicate that most Russians support the war.
As always, thank you to all new subscribers to this newsletter, and my various inboxes are open for feedback, suggestions, tips, leaks, ideas and gossip.
See you next week,
Next Trend Realty LLC./wwwHar.com/Chester-Swanson/agent_cbswan
1 年Thanks for Sharing.