Misinformation, Fakes & LLM's

Why they matter

I attended a Turing Institute talk given by Alicja Martinek  on LLM’s entitled ‘Deep Fakes and Misinformation’, In July. I found it so fascinating that I felt the wider audience should be aware rather than just academics. I have reproduced most of her observations but before I do that let me make a few observations myself, both as economics and keen historian. The size is huge.

I am a keen believer in the power of AI to improve lives and freedom of social media to inform, but the Wild West of needs urgent regulation and strict governance before we have data dictatorships.

With 5.31 billion social media users worldwide – accounting for 63.9% of the global population – social platforms have become necessary hubs for gathering information, connecting with our friends and loved ones, and growing businesses.

Data landscape

Social Media Platform   Monthly Active Users (MAU)

1          Facebook          3.07 billion

2          YouTube           2.5 billion

3          Instagram          2 billion

4          TikTok              1.6 billion

5          Snapchat           850 million

6          X (Twitter)        586 million

7          Pinterest           570 million

8          LinkedIn           350 million

9          Threads             275 million

10        Reddit              185–226 million

That is one hell of reach of social media.

How much information is true or false?

Misinformation is nothing new. It as old as recorded history and history is often distorted to suit the narratives of the story teller or the rulers. This stems all the way back from the Nebucanesser to the Egyptians, Greeks and modern spin doctors. This more common in religion which demands certain gullibility, suspension of common sense and need to belong by followers. It is common in politics where politicians pander to the voter’s base instincts of fear, hate and envy.

In this heavy mix, misinformation or lies are the common currency of power and influence. Add the impact of what Vance Packard called the ‘Hidden Persuaders’, consumerism and powers to influence with the digital social media and you have perfect recipe for power of fake news and misinformation campaigns to impact voting/buying behaviours, power and social and political and global stability.

For example, Wikipedia published that fact-checkers at The Washington Post documented 30,573 false or misleading claims during his first presidential term, an average of 21 per day. Scholarly analysis of Trump's X posts found significant evidence of an intent to deceive. The purpose is to:

1.      Flood the channels with  over information

2.      Lie till falsehoods become facts in peoples’ minds

3.      Lie to cause confusion and doubt as to what is fact and what is fiction.

Social misinformation channels

For an ordinary informed person it is difficult enough to distinguish fact and fiction but you can see how the half educated and gullible or greedy, with promises of salvation or wealth, can be persuaded to vote for their own demise. Dr Martinek explained how misinformation is used.

Spread false information to:

1.      Cause panic, polarise a nation or destabilise it.

2.      Fake mass events-inciting action

3.      Create fear

4.      Impact elections-call to voters to cast a vote or not.

5.      Impact state security-Ukraine, Indo Pakistan, Israel and Hamas

6.      Fishing for confidential information (bank details, love scams) etc

7.      Threats and military threats

8.      Fraudsters/Hackers

Building tensions

This is used to cast doubt or create call for action and sell ideas or goods

1.      Using time pressure or emotions

2.      Pretend to expose conspiracies

3.      Sensationalise

4.      Repetition

5.      Superlatives

6.      Discredit others

7.      Competitiveness

8.      Transparency-trust us

9.      Ease of reaching your goals with us

10.   Offer luxury and limited products

11.   Minimise risks by taking action now

12.   Give a promise –we guarantee …..

13.   Call to action-‘press now’

14.   Altruism- call to action to help others, fight poverty, water aid etc

15.   Targeting normal people’s needs

16.   Address people genuine concerns

17.   Social endorsements

These are just some of the ways social media and AI is used to get us to take or refrain from action but the defence is the very tool used to spread false ideas or misinformation.

Defence using AI

1.      Most social media platforms have tools to detect threats or what might me false

2.      Tools to detect synthetic speech. Chat-bots

3.      Images and video can be detected as false

4.      False claims in medical or by Parma out of 153 claims 125 were false

Large language Model (LLM’s)

LMs, or Large Language Models, are AI systems trained using  massive amounts of online text data to understand and generate human-like text. Google and others have the large data sets that help use simply or clarify tasks. They can perform various natural language processing tasks, such as translation, summarization, and question answering. Like child, the larger its vocabulary the better it can construct its arguments or response.

But as philosopher Ludwig Wittgenstein said: ‘Whereof one cannot speak, thereof one must be silent.’ As quoted in ‘Tractatus Logico-Philosophicus. In other words the language we use is the limit of our world.  Let’s do a thought experiment. If you image being in one room, with no window or door but high wall and a ceiling. We would know no other world. This is our world of ‘facts’.

Our world would be constrained by our language, our senses and perception, the so called facts of our world. If some said there is world outside this room we would have to deny it. This is not our world. Logically and rationally this is what LLMs are creating. Most social media is Eurocentric as is the AI being developed from it. Not very healthy given its past.

Two issues for AI

1.      If LLMs and AI merely reconfigures data , is it really creative or merely innovative

2.      Do LLM’s limit growth of knowledge?

I feel this is for another article but as human development never stops, except in the world of religious extremist, how to do we help people distinguish fact from fiction. We have mentioned a few points.

Education reforms

Rather than teaching for Victorian age that does not exist, except in our educational system we have to radically change education system and teach to be life practically orientated.

1.      Critical thinking- teaching the basics or logic and philosophy so the nest generation obtain abilities to distinguish right and wrong.

2.      Historical thinking so they learn how to investigate events to distinguish facts from fiction and hearsay.

3.      AI for good code, like Hippocratic for doctors, which often refers to the Hippocratic Oath, a set of ethical principles traditionally associated with physicians. There is no reason why a similar one cannot be done and certificates issues for transparency and veracity for practitioners who apply the principles.

4.      The educational levels from poor (Religious societies and states) to high (none religious states) to be improved to ensure that charlatans, like Trump or religious bigots, are fact checks and if certain number of claims are made  which are lies or untruth they are banned for life.

5.      Democratic rethink on how it works and the processes to ensure that democracy does not become a tool for dictatorship.

Conclusion

Unless we take radical steps without critical thinking in people, using LLM, in social media will lead to social and political instability. The raise of false leaders supported by big business as state and none state players get better at using AI for  misinformation.  We have to take action now otherwise AI will be corrupted as the UN or the US political system with misinformation and policies missteps. Thanks to Turing Institute to making are rethink my position on key issues.