Dr Simon Thorne

Programme chair for The European Spreadsheet Risks Interest Group, EuSpRIG - (“yewsprig”) for short. Academic and researcher in software and spreadsheet quality

[Dr Simon Thorne @EuSpRIG]

It’s really happening! EuSpRIG welcomes the one and only @standupmaths. We’re ready for risks, research, and possibly the most entertaining spreadsheet jokes ever committed to cell. SUM-mon your curiosity. #EuSpRIG2025

https://t.co/gx2ximNB8o

[Matt Parker @standupmaths]

It’s finally happening! I will be at the European Spreadsheet
Risks Interest Group annual conference! See sum(you all) there.

[Dr Simon Thorne @EuSpRIG]

@owenjonesjourno Well I shall be on BBc News this evening at 6pm to talk about AI safety!

[Dr Simon Thorne @EuSpRIG]

@owenjonesjourno This is exactly what they are trying to make you think of. Honestly these claims around AI are baseless nonsense that are designed to hype up the capabilities of LLMs and keep the cash flowing. Those assertions made by the ex anthropic employee are science fiction not fact.

[Dr Simon Thorne @EuSpRIG]

@freethepolitics @owenjonesjourno ... use it to generate non consential sexualised images, because it hallucinates, because it is biased, it negates the education system with short cuts, it is a privacy risk, it can be used to hack computer systems and all sorts of HUMAN centric problems.

[Dr Simon Thorne @EuSpRIG]

@freethepolitics @owenjonesjourno There are legitimate risks that arise from generative AI and LLMs but these are HUMAN problems not AI. I.e. it needs regulating and at present it is totally unregulated in the UK. That needs to change but not because AI will wipe out humanity but because HUMANs will....

[Dr Simon Thorne @EuSpRIG]

@freethepolitics @owenjonesjourno And can ASI be reached? No, not with current LLM technology. It is way too limited. They will say otherwise to keep the cash flowing. Look at how much money they are losing, they will say anything to keep going.

[Dr Simon Thorne @EuSpRIG]

@freethepolitics @owenjonesjourno They did hack in but not "unprovoked" - essentially the company didn't secure the environment the LLM was placed in meaning it could access hugging face and hack it. It is human error it is not evidence of some AI entity turning on its creators. Do not believe the BS

[Dr Simon Thorne @EuSpRIG]

@Channel4News No it’s total nonsense, illiterate bollocks.

[Dr Simon Thorne @EuSpRIG]

@BernieSanders @BernieSanders it’s nonsense! This is simply keeping it in the news. LLMs cannot produce this kind of outcome. It’s all smoke and mirrors

[Dr Simon Thorne @EuSpRIG]

@SkyNews Don’t get sucked into this baseless nonsense, click bait

[Dr Simon Thorne @EuSpRIG]

@BBCNews Total bullshit! This illiterate nonsense and you should know better @BBCNews

[Dr Simon Thorne @EuSpRIG]

@owenjonesjourno This post from Jacob is nonsense, don’t get swept up in the bullshit.

[Dr Simon Thorne @EuSpRIG]

@hilbertspaess Illiterate nonsense I’m afraid

[Dr Simon Thorne @EuSpRIG]

@demishassabis Total fantasy not grounded in reality read this: https://t.co/HxLzvlKJer

[Dr Simon Thorne @EuSpRIG]

Probably the biggest most impactful spreadsheet error ever made...

[Hongshen Zhu @HongshenZhu]

Indeed a crazy story. “High debt share hampers GDP growth” is still somehow a received wisdom preached to developing countries including China, yet just a boomer fat finger error.

[Dr Simon Thorne @EuSpRIG]

@nicochristie @BrianSJ3 I wonder though would an experienced professional want to use it at all? i.e. someone who knows what they are doing would probably not bother to AI generate things. The auditing aspect of it I like a lot though, I have a much harder more abstract task which I will test it with.

[Dr Simon Thorne @EuSpRIG]

@BrianSJ3 @nicochristie I suppose the question is who is this aimed at exactly... the generation stuff was quite interesting too, it was able to create functions from plain English but one would assume then the user doesn't know how to do this. I am not sure it saves time if you know what your doing.

[Dr Simon Thorne @EuSpRIG]

@BrianSJ3 @nicochristie I have done some elementary testing with this, in my basic error seeded workbook, it identified 11 out of 12 issues which is quite impressive and gave a good explanation of what was wrong and how to fix it.

[Dr Simon Thorne @EuSpRIG]

@BrianSJ3 @nicochristie Interesting, I wonder how robust that is?

[Dr Simon Thorne @EuSpRIG]

Spreadsheet - a response to an FOI request. Again if you’re a malicious actor then that list of names could mean they are able to target the names on the list. All because we forgot to check something properly before we rush it out the door! Despite all the security elsewhere

[Dr Simon Thorne @EuSpRIG]

This is correct, it was done in good faith - the spreadsheet had the names of some trusted individuals, people who could be vouched for - I.e. they maybe worked with the MOD in Afgan - interpreters, intelligence, local community liaison etc. The spreadsheet…

[Respect @rrespectorr]

@en4rab @EuSpRIG I think they actually shared it accidentally in an email to unrelated people and that's how it got leaked

[Dr Simon Thorne @EuSpRIG]

active police units, CT units, MOD staff and MI5 staff working with the police. Some of which of which were on active duty at the time. The information included addresses of work places, names, ranks all sorts of highly sensitive information in a hidden worksheet in the…