The Conversation
Quote of the Day
“The war against intelligence is always waged in the name of common sense.”
Roland Barthes
Musical alternative to the morning’s radio news
Tuba Skinny | Jubilee Stomp | Royal Street
Long Read of the Day
How to interact intelligently with a Large Language Model
As recounted by Terence Tao on his blog.
Terence Tao is a world-class mathematician. (He won the Fields medal — math’s equivalent of a Nobel prize — in 2006. He also serves on the US President’s Council of Advisers on Science and Technology.) This blog post is a fascinating example of how one might make good use of a resource like CPT-4.
Here’s the process he followed.
First he asked GPT-4 to answer the questions of how might AI technology and its successors contribute to human flourishing, and how might we as society best guide the technology to achieve maximal benefits for humanity – and then to generate an essay that he might write on this topic.
While the machine was thinking, he wrote the first draft of the article himself.
Then he gave GPT-4 his essay and asked it to rewrite it to more closely resemble his personal style.
Finally, he asked the model to directly improve the writing of his own article to make it more effective.
The post reproduces all the essays involved in this interaction.
I found it fascinating. Hope you do too.
Chatbots: social media on steroids
My column in yesterday’s Observer:
The other thing about chatbots is they enable the effortless creation of massive quantities of “content” on an extraordinary scale. As James Vincent of the Verge puts it, “Given money and compute, AI systems – particularly the generative models currently in vogue – scale effortlessly. They produce text and images in abundance, and soon, music and video, too. Their output can potentially overrun or outcompete the platforms we rely on for news, information and entertainment. But the quality of these systems is often poor, and they’re built in a way that is parasitical on the web today. These models are trained on strata of data laid down during the last web age, which they recreate imperfectly.”
Soon, though, the web might consist not only of what was there in the pre-AI era, but all the stuff created by current and future chatbots. Which raises the intriguing possibility of an online world populated by bots inhaling the textual exhaust of their mechanical peers, and a consequent spiral into the infinite recursion that programmers call “stack overflow”!
In such circumstances, what should truth-seeking institutions do? Answer: look at what they are doing at Wikipedia…
Do read the entire piece
Moral depravity, UK style
From the morning edition of the FT newsletter the other day.
The British government’s Rwanda policy continues to be a great piece of statecraft: by Paul Kagame, that is. He has essentially bought the government’s Africa policy with £120mn of the UK’s own money — paid by the British government to the Rwandan one — before a single deportation flight has left the UK for the African nation. He can look forward to much more money if — though it is a very big “if” — the UK government ever manages to implement the policy. It will seek permission to appeal against the Court of Appeal’s latest ruling at the Supreme Court.
Last week the US and the EU called on Rwanda to cease its alleged support for M23, the militia that re-emerged in 2021 to wage an offensive in the eastern Democratic Republic of Congo. The US, EU and the DRC all say the group is backed by Kagame’s government. … But, because of the deal struck with Kigali, the UK has said nothing at all.
Last week the UK Court of Appeals ruled that the government’s scheme to deport asylum seekers to Rwanda is unlawful, on the grounds that Rwanda is not a safe third country.
Which it isn’t.
My commonplace booklet
End of the (paper) Line
This is the front page of the Wiener Zeitung on Friday, the last day it appeared as a printed newspaper. In it the Editor penned a farewell letter (in German) which Google Translate rendered thus:
Dear readers, as I say goodbye, I would like to invite you to attend my last journey, which finally brought me to you in printed form. Delivered for the last time because even a newspaper owned by the Republic of Austria has to admit that from time to time you are at the mercy of politicians who have decided to only publish me digitally in the future. Robbed of my outfit, from tomorrow the haptics will no longer be determined by paper, but exclusively by swiping on the smartphone and clicking on the screen with the mouse. My future inner values, i.e. the content that the largely newly composed editorial team will only deliver digitally from tomorrow? You will see it if you want. I don’t know it.”
Linkblog
Some things I noticed while trying to drink from the Internet firehose.
Embracing change and resetting expectations. Those of us who work with computers grew up taking as an article of faith that every instruction we gave to a machine had to be precise in every pedantic detail. The advent of tools like LLMs, argues Terence Tao, means that these assumptions will need to be recalibrated, if not abandoned entirely.
Carriers plan to rescue a few more unused smartphones • The Register. The GSM Association reckons that five billion mobile phones are “currently sitting unused in desk drawers around the globe”. Their innards contain 50,000 tonnes of copper, 500 tonnes of silver and 100 tonnes of gold. There’s also enough cobalt to build batteries for ten million electric vehicles. The Register’s correspondent is “puzzled by this talk of desk drawers”. His dead mobiles are scattered among “a rather nice wicker picnic basket, a filing cabinet, and That Box Full Of Old Tech I Should Probably Have Thrown Out But Kept Just In Case”. Me too.