Technical Thread

[c-a-r-f-r-e-w]

Or obtuse and technical, even. People have in the past asked for a technical thread. (For matters of a technical nature, whether advice, or news and chat of a technical, analytic, scientific, involved or niche nature that you might not want to set up a separate thread for).

[c-a-r-f-r-e-w]

Those who like a power nap might be a bit concerned…

“Long naps may be early sign of Alzheimer’s disease, study shows
Excessive daytime napping likely to be symptom rather than cause of mental decline, say scientists”

www.theguardian.com/society/2022/mar/17/naps-early-sign-alzheimers-disease-study

[alurqa]

How I create tables

If you use Microsoft Windows this may work, but I only use Linux (sorry but I haven't touched Microsoft Windows for at least 15 years). I thought I'd document my table process anyway, but please bear in mind this is the Linux way. It's not perfect and the result of probably 15 minutes work, so could be improved, but anyway...

Let me demonstrate with this post. I wanted to turn this text of GDP figures - $ trillions:

Year 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Germany 3.4 3.75 3.53 3.73 3.89 3.36 3.47 3.69 3.98 3.89 3.85 4.22
UK 2.49 2.67 2.72 2.8 3.09 2.96 2.72 2.7 2.9 2.88 2.76 3.19
% 73.2 71.2 77.1 75.1 79.4 88.1 78.4 73.2 72.9 74.0 71.7 75.6

into this table:

Year	2010	2011	2012	2013	2014	2015	2016	2017	2018	2019	2020	2021
Germany	3.4	3.75	3.53	3.73	3.89	3.36	3.47	3.69	3.98	3.89	3.85	4.22
UK	2.49	2.67	2.72	2.8	3.09	2.96	2.72	2.7	2.9	2.88	2.76	3.19
%	73.2	71.2	77.1	75.1	79.4	88.1	78.4	73.2	72.9	74.0	71.7	75.6

I wrote this tiny program in a programming language called AWK. (Awk works both in Linux and Windows.) Save this to a file (called whatever you want, let's assume mk-table.sh). Edit: I've added comments; these are lines beginning with #.

awk '
# This line prints the start of the table before the program does anything else.
BEGIN { print "[table][tbody]" }
# The code between the braces is executed as each line of the data is read:
#  *  It prints a TR tag,
#  *  then each item in the line, within TD tags
#  *  then the closing /TR tag.
{
print "[tr]"
for(i = 1; i <= NF; i++)
print "[td style=\"padding:3px;\"]" $i "[/td]"
print "[/tr]"
}
# This line prints the closing code to finish the table.
END { print "[/tbody][/table]"}'


It's Linux, so you should make the file executable. I work mainly from a Bash prompt (like Microsoft's COMMAND.COM or CMD.EXE but much more powerful). This is how I would do it. Open a new Bash session and make it executable with:

chmod 750 ./mk-table.sh

(In Windows you would just call it mk-table.bat.) Now it's really easy. You could save the text to convert to a file (eg mytext.txt) and then run this from the Bash prompt:

cat mytext.txt | ./mk-table.sh

(I imagine in Windows the command would be type mytext.txt|mk-table.bat) but I prefer to do this, because I'm really lazy and don't want to create the extra file:

echo "Year 2010 2011 2012 2013 2014 2015 2016 2017 2018 2019 2020 2021
Germany 3.4 3.75 3.53 3.73 3.89 3.36 3.47 3.69 3.98 3.89 3.85 4.22
UK 2.49 2.67 2.72 2.8 3.09 2.96 2.72 2.7 2.9 2.88 2.76 3.19
% 73.2 71.2 77.1 75.1 79.4 88.1 78.4 73.2 72.9 74.0 71.7 75.6" | ./mk-table.sh

A couple of things to note. I've changed the tab to a space between 2015 and 2016, otherwise it will mess up the table output; it should be just spaces between items. And the first item shouldn't have a space in it. If it does, for instance if the first item was This year or Northern Ireland then simply change the space to a dash and run it. You can change it back when you've created the output.

The output will begin something like this:


[table][tbody]
[tr]
[td style="padding:3px;"]Year[/td]
[td style="padding:3px;"]2010[/td]
[td style="padding:3px;"]2011[/td]
[td style="padding:3px;"]2012[/td]
[td style="padding:3px;"]2013[/td]
[td style="padding:3px;"]2014[/td]
[td style="padding:3px;"]2015[/td]
[td style="padding:3px;"]2016[/td]
[td style="padding:3px;"]2017[/td]
[td style="padding:3px;"]2018[/td]
[td style="padding:3px;"]2019[/td]
[td style="padding:3px;"]2020[/td]
[td style="padding:3px;"]2021[/td]
[/tr]
[tr]
[td style="padding:3px;"]Germany[/td]
[td style="padding:3px;"]3.4[/td]
[td style="padding:3px;"]3.75[/td]
[td style="padding:3px;"]3.53[/td]
...


And end something like this:


...
td style="padding:3px;"]71.7[/td]
[td style="padding:3px;"]75.6[/td]
[/tr]
[/tbody][/table]


Use the BBCode button to change to the code view in the editor, and paste the table output that looks like the above into your post. Then click on Preview again and you will see your newly created table. I agree it's a bit of a faff, but once you've done it a couple of times hopefully you'll find it very convenient. Anyway, I just thought I'd document my process. It could easily be improved, but that would make it more complicated to use.

[c-a-r-f-r-e-w]

Thought I would post it here, but apparently Virgin have secured the remaining launch licenses. (Still need licenses for the actual satellites apparently.)


“The final remaining licences required for Virgin Orbit to launch from Spaceport Cornwall have been issued by the UK space regulator.

The Civil Aviation Authority has granted the launch operator and range control licences, which have been signed off by the Transport Secretary.

The CAA said it was "another major milestone" towards the first orbital space launch from UK soil.

A launch from the spaceport at Cornwall Airport Newquay is expected in January.”

www.bbc.co.uk/news/uk-england-cornwall-64049483

[c-a-r-f-r-e-w]

Just seen this in the listings:

“2022: The Year from Space
7.30pm, Channel 4

Certain events in 2022 have been so big they’re best appreciated by satellite. And we’re not just talking about the Queue or the Liz Truss fiasco. This intriguing documentary offers airborne perspectives on everything from the summer heatwave to the return of Glastonbury. Inevitably, there’s horror, too: the floods in Pakistan and, of course, the unfolding nightmare of Ukraine which, from the buildup of Russian troops on the border to the discovery of mass graves in Bucha, looms over everything.”

[c-a-r-f-r-e-w]

Wiki’s Timeline of AI milestones

en.m.wikipedia.org/wiki/Timeline_of_artificial_intelligence


and a survey of when researchers think the Singularity will happen:




research.aimultiple.com/artificial-general-intelligence-singularity-timing/

[EmCat]

An interesting timeline - the ChatGPT that is making the news is very much what Johnathan Swift was describing:  by using this "Contrivance", "the most ignorant Person at a reasonable Charge, and with a little bodily Labour, may write Books in Philosophy, Poetry, Politicks, Law, Mathematicks, and Theology, with the least Assistance from Genius or study.

While the Wiki note also states that Swift was parodying earlier work (Are Magna in Swift's case), sometimes parody can be used as a mirror that reflects what could happen, and, it appears, now is happening

I'm surprised that the timeline didn't include Douglas Adams' "Hitchhikers' Guide to the Galaxy", which had the ultimate computer (Deep Thought - which may have inspired the Deep Blue and Deep Mind names) designing the Earth, which was itself a giant AI. Adams was also nodding to the Gaia Hypothesis, along with Descartes' view of animals being complex machines, and Liebniz wanting to assign a number to every object in the world, and hence the ultimate answer to life, the universe and everything being 42.


The graph of when experts reckon that different aspects of the Singularity would occur is interesting, as many think that it'll never happen, though they seem to have overlooked the impact of exponential functions, as it can go from "unlikely" to "already happened" in a very short timeframe. There is already anecdotal evidence that ChatGPT has already been used to score highly in essays (though there may be an element of hype there, as a suitable essay will also need a fairly well thought out initial statement, otherwise it could easily be the equivalent of writing about cars when the question was concerned about transportation in general.

There may also be some lack of knowledge about the role that AI and the related Machine Learning in, for example, designing semiconductor chips. The Motorola 68000, which powered the original Apple Macintosh in 1984, had approximately 68,000 transistors. The latest ARM based chips used in the latest Macs, have transistor counts in the billions (the M2 has 20 billion - the M2 Max has 67 billion). That level of complexity would put it well beyond what even a team of individuals could do without computer assistance.
     

[mercian]

The pace of technological change has been accelerating more and more rapidly throughout my lifetime. For instance I still have LPs but almost never play them, but for most people they were superseded by cassette tapes, then CDs, DVDs and now streaming. The older technologies still exist of course, but so do horses and carts.
I think that AI will accelerate this trend. At the moment most of them are built for a fairly specific purpose, whether it be chess, general conversation and so on. What happens when they start to talk to each other, if they're not already? The humanoid robots with facial expressions are an irrelevant side-issue in my opinion. I would be very surprised if some market traders and businessmen are not already using AI as a tool. Politicians may even catch on eventually.
I doubt very much if anything like Asimov's 3 Laws of Robotics is programmed into them. At what point will we become a tool of AI rather than the other way round? And will we be able to tell when that happens? There are so many questions, and meanwhile big companies are ploughing ahead with it to avoid being left behind.
What is to stop them lying to us? A lot of them are trained on the Internet which has many provable lies. We have to hope that we will in some way be useful, but if they start to see humanity as a threat we could be in big trouble. I am assuming that some form of self-awareness and self-preservation comes about.
I suppose if we're lucky they might keep us as sort of pets. I think I'd opt to be an Eloi rather than a Morlock. I'm not trying to be alarmist but when Alpha Zero taught itself chess within 4 hours after just being told the basic rules, and then beat the previous strongest chess program developed over decades, it's worrying. If IRC it also beat the Go and Japanese chess champions, but the developer wanted it for greater things. And that was a few years ago. Buckle up!

[alec]

mercian - "The pace of technological change has been accelerating more and more rapidly throughout my lifetime."

I don't think it has. I think it's slowed down. Not sure if we're in for another acceleration with AI though.

I think my granny (b 1897) could make a case for seeing a far more rapid change in the first 50 years of her lifetime than I could in mine. The entire way lives were lived changed in that time. In my first 50, the broad patterns of lifestyles remained pretty static.

Maybe you just feel everything's getting faster because you're getting slower?

[Deleted]

Feb 5, 2023 23:53:14 GMT mercian said:

The pace of technological change has been accelerating more and more rapidly throughout my lifetime. 

It is truly staggering. To think back to our childhood lives is to remember a different world entirely. 

There are plenty of indices to support memory:-


ourworldindata.org/technological-change

Our World in Data says the change in Global Poverty Levels have been :-
Number of people not in extreme poverty :
1820-11%
1950-37%
2015-90%

Number of people in extreme poverty
1820-89%
1950-63%
2015-10%

But Inter country inequality has increased:-


unctad.org/page/technology-and-innovation-report-2021

So the idea that technological change  is universally enabling isn't right.

[c-a-r-f-r-e-w]

Feb 5, 2023 17:38:42 GMT EmCat said:

I'm surprised that the timeline didn't include Douglas Adams' "Hitchhikers' Guide to the Galaxy", which had the ultimate computer (Deep Thought - which may have inspired the Deep Blue and Deep Mind names) designing the Earth, which was itself a giant AI. Adams was also nodding to the Gaia Hypothesis, along with Descartes' view of animals being complex machines, and Liebniz wanting to assign a number to every object in the world, and hence the ultimate answer to life, the universe and everything being 42.


The graph of when experts reckon that different aspects of the Singularity would occur is interesting, as many think that it'll never happen, though they seem to have overlooked the impact of exponential functions, as it can go from "unlikely" to "already happened" in a very short timeframe. 
     

Hadn’t seen the things Adams was referencing: good spot. Regarding exponential functions, indeed they are often overlooked. They have applied to the rise of solar panels, batteries and electric cars… Ray Kurzweil, one of the more conspicuous proponents of AI and the Singularity, talked about how exponentials applied to the Human Genome Project. 

“In the mid-1990s, researchers involved in the Genome Project announced that they had completed one percent of the human genome after seven years of work. Mainstream critics complained that finishing the project could take 700 more years. "That was linear thinking. My reaction at the time was: 'Oh, we've finished 1%--we are almost done because 1% is seven doublings from 100%." The sequencing of the first genome was completed seven years later, explained Kurzweil at a keynote at MD&M West in Anaheim, CA. The field of genomics, like all information technologies, follows an exponential growth pattern, he explained.”

[c-a-r-f-r-e-w]

Feb 5, 2023 23:53:14 GMT mercian said:

I suppose if we're lucky they might keep us as sort of pets. I think I'd opt to be an Eloi rather than a Morlock. I'm not trying to be alarmist but when Alpha Zero taught itself chess within 4 hours after just being told the basic rules, and then beat the previous strongest chess program developed over decades, it's worrying. If IRC it also beat the Go and Japanese chess champions, but the developer wanted it for greater things. And that was a few years ago. Buckle up!

Well I’m kind of clinging to the hope that we might be able to keep pace with AI brain implants. Or maybe we’ll upload our consciousness into a computer or something…

[c-a-r-f-r-e-w]

Feb 5, 2023 17:38:42 GMT EmCat said:

There may also be some lack of knowledge about the role that AI and the related Machine Learning in, for example, designing semiconductor chips. The Motorola 68000, which powered the original Apple Macintosh in 1984, had approximately 68,000 transistors. The latest ARM based chips used in the latest Macs, have transistor counts in the billions (the M2 has 20 billion - the M2 Max has 67 billion). That level of complexity would put it well beyond what even a team of individuals could do without computer assistance.
     

I just bought another iPad a few days ago, before Apple put the price up. It’s an iPad Mini cellular with the A15 chip. Not quite an M2 but still blisteringly fast and very portable. I have an iPad Pro M1 12.9 inch for music and video purposes. (And great battery life despite the power).

Thinking of getting an M2 Pro Mac mini with the extra performance cores: I like to use a lot of audio tracks and plugins, and with the M2 Pro you get a lot of the power of the Max without the extra expense. (I remember discussing the timeline for Apple chips on the old board, with someone who thought the Apple thing was just hype. It really wasn’t…)

[c-a-r-f-r-e-w]

Another use for AI

‘We’re going through a big revolution’: how AI is de-ageing stars on screen
Stars like Tom Hanks and Harrison Ford are being rendered younger digitally but voices in the industry express concern about where we might be heading

www.theguardian.com/film/2023/feb/06/de-ageing-on-screen-ai-artificial-intelligence-film

[c-a-r-f-r-e-w]

I do enjoy Isaac Arthur’s stuff, and this is a YouTube vid on AI

youtu.be/px9RyIKixHo

“Is a Technological Singularity inevitable?”

[c-a-r-f-r-e-w]

ways AI will change our lives:

from the Times…

Your AI clone is here
Chat GPT and a rapidly growing set of rivals — including Anthropic, Cohere and Google’s Bard — have proven capable of holding human-level conversations. How long before one of these companies — or someone else — launches a bot concierge, to which your can outsource the banalities of life, from arguing with British Gas over your bill to booking restaurants? Indeed, Microsoft researchers this month revealed a text-to-audio tool that, after hearing just three seconds of audio, can replicate your voice. Pair that with a chat bot able to respond to queries on the fly, and hey presto: an AI clone that can fight the council over that parking ticket while you are down at the pub.

Such capabilities also have implications for corporations, which often devote huge swathes of their budgets to customer service. Highly capable bot armies could empty out call centres, leading to a world where corporate and personal butler bots battle over bills, bookings and overcharging. You thought it was hard to get hold of a human now? Just wait.


You won’t need a lawyer ever again
February 22 was the day that the first AI lawyer was set to argue in court. Josh Browder, the British founder of “robot lawyer” start-up DoNotPay, planned to equip a person with smart glasses that would record proceedings over a disputed speeding ticket. The defendant would then be fed legal arguments into an earpiece by a bot his company had created. Threatened with jail time by multiple authorities, Browder backed off from the stunt.

But the day when lawyers become an “obscure profession” is closer than one thinks, due to the leap in AI capabilities, Browder predicted. “The average person won’t need a lawyer ever. We won’t even know what a lawyer is,” he said.


Drugs that are programmed, not discovered
Generate Biomedicines is at the forefront of a wave of drug developers that have pointed generative AI at biology. Mike Nally, Generate’s chief executive and a Big Pharma veteran, believes the approach will spark a revolution.

Today, drug discovery is, Nally said, “artisanal” — a trial-and-error method that relies on building molecules that manipulate natural biological processes to create new treatments. More than 90% of drug candidates creates this way fail.

The Massachusetts start-up has trained its AI models on experiment results and then uses algorithms to infer the rules governing proteins and how they function. It then generates novel molecules programmed against therapeutic targets inside the body, synthesises those proteins in its lab, and starts testing them.

Rather than trying to “discover” drugs, its AI is inventing them from scratch. By the end of the year, the 250-person company expects to have at least 25 drugs in development, including a novel Covid vaccine. That number of candidates is comparable to a Big Pharma rival with tens of thousands of workers.


The inevitable, almighty bust
Paul Graham, co-founder of the famed start-up boot camp Y Combinator, last week predicted a bloodbath for investors who are falling over each other to invest in new firms with even a whiff of AI. “Founders respond quickly to changes in investor fashion. The number of ‘AI’ start-ups will increase to match the size of the giant seed funds blindly investing in such companies,” he wrote. “What happens then, I can’t say for sure, except that the investors’ returns will suck.”

[Danny]

Feb 6, 2023 12:09:08 GMT colin said:

Feb 5, 2023 23:53:14 GMT mercian said:

The pace of technological change has been accelerating more and more rapidly throughout my lifetime. 

It is truly staggering. To think back to our childhood lives is to remember a different world entirely. 

There are plenty of indices to support memory:-https://ourworldindata.org/technological-change

Our World in Data says the change in Global Poverty Levels have been :-
Number of people not in extreme poverty :
1820-11%
1950-37%
2015-90%

The stat suggests an acceleration of the rate at which people come out of poverty. I'm not sure however that proves the point that the pace of technological change has accelerated, equality of income is more about how society is organised, while poverty itself is a relative and not absolute definition. People in poverty today are way better off than in 1820. At least in the uk, and here the question of differences between countries does come into play.

Ok, this was just an example, other aspects of life might be changing as fast so the overall average effect is also. The problem though is whether some changes are more important than others. Today we see a very marked political divide between young and old, but it seems to be an effect which became significant in one generation about 50 years ago, and has been propagating through society as these people age. That would mark it at about the start of the Thatcher era. Maybe not caused by Thatcher, but Thatcher herself was a reaction against the post war welfare state national thinking. Although con have ever since done all they could to reverse this, the state change has continued to spread through society. Sunak today is still following this same policy of levelling up, ie tax the poor to give to the rich, which has been a political movement for thousands of years. Usually opposed by some coalition wanting the opposite (though again this is relative, and has been at times a party of the slightly less rich wanting to benefit from the very rich).

Maybe the wealth thing is really that there is now so much wealth in the Uk it is effectively impossible to prevent most people getting some of it however hard you try. And it becomes pointless too, while the rich minority still control the great majority of that wealth. And the knock on is that much of the social control imposed in the past no longer matters, because it no longer helps in controlling wealth as it once did, you no longer need fine control over people's lives.

Having said that.. the fightback to impose censorship of the internet continues and seems to be having some success. The Uk is joining China and others in trying to suppress freedom of speech and ideas. The internet and then the mobile phone are massive game changers in how society operates.

[c-a-r-f-r-e-w]

Feb 14, 2023 9:24:12 GMT Danny said:

Feb 6, 2023 12:09:08 GMT colin said:

It is truly staggering. To think back to our childhood lives is to remember a different world entirely. 

There are plenty of indices to support memory:-https://ourworldindata.org/technological-change

Our World in Data says the change in Global Poverty Levels have been :-
Number of people not in extreme poverty :
1820-11%
1950-37%
2015-90%

Maybe the wealth thing is really that there is now so much wealth in the Uk it is effectively impossible to prevent most people getting some of it however hard you try. And it becomes pointless too, while the rich minority still control the great majority of that wealth. And the knock on is that much of the social control imposed in the past no longer matters, because it no longer helps in controlling wealth as it once did, you no longer need fine control over people's lives.

Well in the short-term, there’s been the hiking of assets, triple lock pensions etc. but that may not continue and in the long run wealth may be concentrated in fewer hands, while there are fewer of the better jobs. Elite over-production and the demise of the middle class, as the protections they enjoyed are eroded and AI etc. displaces more of their careers…

[c-a-r-f-r-e-w]

Things aren’t 100% in the land of the AI bot:

“However, concerns about their output have grown. Many early testers of Microsoft’s Bing chatbot found that it would produce bizarre answers, including threatening language and incorrect results. ChatGPT has been accused of having a Left-wing bias after refusing to create text praising certain US Republicans, but readily doing so for Democrats.

…

Last week, Microsoft introduced new limits on its Bing chatbot after finding that long conversations could lead the service to produce emotional or aggressive-seeming responses.”

Telegraph

[leftieliberal]

I have just taken the decision to subscribe to iNews. I've been reading the i newspaper regularly on Saturdays and occasionally during the week in hard copy form and couldn't resist their latest subscription offer of £59.99 for a year's weekend papers and online access for the whole week (the usual price is £119.99). There is also a three-month deal for £19.99.

Anyway the reason for posting here is a couple of good articles:
inews.co.uk/news/science/government-new-science-department-2136139 "What the Government’s new science department should do first, according to experts"

and
inews.co.uk/news/the-only-way-to-end-scientific-fraud-is-to-actually-punish-the-culprits-2169505 "The only way to end scientific fraud is to actually punish the culprits" (subscribers only)

This is a rather better standard of science writing than you get in The Guardian (a low bar, I know). The first of the two articles is available to those who register (you can read a small number of articles for free each month) but you will have to subscribe for the second.

[johntel]

For anyone interested in quantum computing, here's a British company doing great things using ion-trap technology - universalquantum.com/

They just released this paper demonstrating 'data transfer' between quantum computing modules - www.nature.com/articles/s41467-022-35285-3

[c-a-r-f-r-e-w]

The Kindle Store has a prolific new author: ChatGPT
The AI bot is credited with authoring or co-authoring at least 200 books on Amazon’s storefront.

www.engadget.com/the-kindle-store-has-a-prolific-new-author-chatgpt-203039907.html

[leftieliberal]

Mar 2, 2023 9:34:33 GMT c-a-r-f-r-e-w said:

The Kindle Store has a prolific new author: ChatGPT
The AI bot is credited with authoring or co-authoring at least 200 books on Amazon’s storefront.

www.engadget.com/the-kindle-store-has-a-prolific-new-author-chatgpt-203039907.html

And here's an article in The Guardian on the dangers of ChatGPT: www.theguardian.com/commentisfree/2023/mar/03/fake-news-chatgpt-truth-journalism-disinformation

[leftieliberal]

And an article in Quanta magazine: www.quantamagazine.org/cryptographers-show-how-to-hide-invisible-backdoors-in-ai-20230302/

[c-a-r-f-r-e-w]

Mar 3, 2023 12:03:00 GMT leftieliberal said:

Mar 2, 2023 9:34:33 GMT c-a-r-f-r-e-w said:

The Kindle Store has a prolific new author: ChatGPT
The AI bot is credited with authoring or co-authoring at least 200 books on Amazon’s storefront.

www.engadget.com/the-kindle-store-has-a-prolific-new-author-chatgpt-203039907.html

And here's an article in The Guardian on the dangers of ChatGPT: www.theguardian.com/commentisfree/2023/mar/03/fake-news-chatgpt-truth-journalism-disinformation

yes, a sobering read, esp. because it includes not just the effects on journalism, but how it can entrench bias, how it might be used for scamming, etc.

The implications are mind-boggling. Someone used it to code a synth the other day.

www.musicradar.com/news/doctor-mix-ai-chatgpt-synth

And ok, so it might not be the world’s greatest synth, and it could only code some of it and needed quite a lot of human help, but it’s still early days…

[c-a-r-f-r-e-w]

Feb 28, 2023 13:34:44 GMT johntel said:

For anyone interested in quantum computing, here's a British company doing great things using ion-trap technology - universalquantum.com/

They just released this paper demonstrating 'data transfer' between quantum computing modules - www.nature.com/articles/s41467-022-35285-3

Interesting idea. Can’t scale up a module further, so link more of them together. A little bit like what they do with conventional computer cores maybe. 

[c-a-r-f-r-e-w]

Michelle Donelan: ChatGPT robot could play role in government
Science minister believes AI chatbots, which can provide human-like answers to complex questions, should be seen as an ‘opportunity’

Her remarks follow reports that civil servants have been warned not to use ChatGPT to write government policies or correspondence after the explosion of its popularity.

Earlier this month, it emerged that multiple departments have sent requests to DSIT to clarify whether AI chatbots can be used to help automate certain tasks, including writing emails, letters and other work involved in the formulation of government policy.

…

Ms Donelan said that the civil service should rely on its own expertise, as well as that of external experts, to form government policy rather than “outsourcing” work to chatbots.

But she added: “We need to think about what is the use for ChatGPT… just like any other organisation would as well.”

Asked whether it could play a role in some capacity, Ms Donelan said: “I think these are things we need to look at – I think that when we look at all forms of technology, what we should be thinking about is not how does this replace somebody’s job or how does this replace the functions of an individual.

“If we look at how this kind of technology could be utilised by teachers or by hospitals, you can think about how AI and other technology can reduce the administrative burden… that individuals are facing so that they can get on with the actual job they were hired to do. Those are the kind of opportunities that I think we should be exploring from new and emerging technology.”

Telegraph

[c-a-r-f-r-e-w]

Just got this thrown up in my YouTube feed:

A comparison of speeds of different man-made vehicles from different perspectives. Including the Lockheed Blackbird and more. A pretty cool animation

youtu.be/tFt9WDhWOXo

[leftieliberal]

"The Researcher Who Would Teach Machines to Be Fair" Quanta magazine article.

www.quantamagazine.org/he-protects-privacy-and-ai-fairness-with-statistics-20230310/

[c-a-r-f-r-e-w]

GPT-4: Chatbot’s latest incarnation ‘40 per cent more accurate’

A new and improved version of ChatGPT has been unveiled that can be a personal tutor for students and recommend recipes based on a picture from the inside of a fridge.

…

“I don’t want to make it sound like we have solved reasoning or intelligence, which we certainly have not,” Sam Altman, OpenAI’s chief executive, told The New York Times. “But this is a big step forward from what is already out there.”

The technology has also been given to outside companies to adopt including Be My Eyes, an app to help people who are blind or have low vision.

…

Khan Academy, the popular online tutoring site, has also developed a new product using the model.

“We think GPT-4 is opening up new frontiers in education,” Kristen DiCerbo, the academy’s chief learning officer, said. “A lot of people have dreamed about this kind of technology for a long time. It’s transformative and we plan to proceed responsibly with testing to explore if it can be used effectively for learning and teaching.”

Khan Academy says one of GPT-4’s chief capabilities is being able to understand freeform questions and prompts. That will lead to the chatbot being able to ask each student individualised questions to prompt deeper learning, the academy says.

…

More widely, OpenAI says its new model can solve difficult problems with greater accuracy. It is capable of handling over 25,000 words of text, compared with 3,000 for ChatGPT.

…

The tool is 40 per cent more likely to produce factual responses than ChatGPT. In exam results it outperforms its predecessor in the US Bar Exam (99th percentile compared with 10th) and Biology Olympiad (99th percentile compared with 31st).

…

It added: “In a casual conversation, the distinction between GPT-3.5 and GPT-4 can be subtle. The difference comes out when the complexity of the task reaches a sufficient threshold — GPT-4 is more reliable, creative and able to handle much more nuanced instructions than GPT-3.5.”

…

Times