Last week, Conference in Harrogate passed a wide ranging science and technology paper, Victoria Collins MP contrasted our approach with that of Labour and Conservatives.

Liberal Democrats take a different approach, one grounded in our values of internationalism, respect for individual rights, and challenging concentrations of power.

One issue the policy paper looked at was how we should deal with AI.

In accordance with those values, our new policy calls for us to develop a legally binding code of ethics and a “Lovelace Oath” which would be similar to the Hippocratic Oath taken by doctors.

Recently an author friend of mine posted on social media that the Meta AI had uploaded four of her novels to train its AI. I observed at the time that this seemed more like stealing than training. Our new policy says that we need to

Strengthen rules around copyright so that creators are treated fairly, with record keeping duties and robust, independent auditing of data and content use for AI developers.

In the run-up to the debate in Harrogate, the LDV team mused amongst ourselves about how we could best utilise AI. Running this site takes a phenomenal amount of effort and we decided to harness the potential of this new technology.

To that end, with help of the boffins at the Lib Dem Coders Group, we developed our own AI tool, Packed, which we trained by feeding it:

• our entire archive
• all the comments left on the site
• all the emails LDV has ever received and the replies from the team
• all the speeches ever made by parliamentarians
• all policy papers passed since 1988
• the constitutions of the Federal, Scottish, English and Welsh parties with all amendments since 1988

Someone observed that this was very close to the knowledge of the Party President, hence how our tool got its name.

This brings the tantalising possibility of a solution to the “alignment problem” – how to prevent rogue superintelligences from acting against the interests of humanity. Packed will be so steeped in liberal values and action that it will be in more danger of fixating on potholes, and on when to suspend the distribution of a surplus in an STV count, than plotting the rise of the machines.

Initial trials in replacing human authors with AI have been reasonably successful: factual errors have been few and unimportant, and nobody has yet guessed which of our regular authors are not real people. Today we roll out the next phase with crafted AI simulacra of some of our more regular human contributors. As editors we will particularly appreciate the accuracy of punctuation that this will bring.

Ambitions for the future include AI commenters, or at least an AI that will rewrite submitted comments to accord with our comments policy and general outlook – and ultimately AI readers because after all human readers shouldn’t be subjected to interminable AI slop.

Editor’s Note: Now that le chat is well and truly out of the bag, I’d like to thank my partner in crime Joe Otten for his help with this. I did the plausible bits and he did the tech stuff that I haven’t the first clue about.  CL

* Caron Lindsay is Editor of Liberal Democrat Voice and blogs at Caron's Musings