Candidates have always been ready to play fast and loose with reality, only too happy to accuse their opponents of having lost their grip on it. Now, however, the reality stakes have suddenly got a lot higher than the true cost of promises. Artificial intelligence is robbing the electorate of the ability to determine what is real and what is not.
Last week an attack ad was released by the right-wing lobby group, the Taxpayers’ Union, depicting Labour leader Chris Hipkins as a home-invading thief. It showed him kicking down doors, frightening an elderly woman, shocking children as he disrupts a family dinner, and stealing fistfuls of cash.
NewstalkZB’s Heather de Plessis-Allan, a steadfast observer of Keep Right signs, thought the ad was “brilliant”. Her Canterbury NewstalkZB stablemate John MacDonald, on the other hand, found the ad “disturbing” because it was authentic enough to trigger adverse psychological reactions in anyone who had been subjected to a home invasion. Their NZME colleague, Shayne Currie, wrote in his Media Insider column that the ad was “part of an increasingly personal and somewhat hostile election campaign”. However, he said on Ryan Bridge Today that, of course, it was labelled as AI and was “plainly satire”. Across on RNZ, political scientist Claire Robinson told Guyon Espiner it was “confronting” and clearly was not satire because it was devoid of humour and commentary.
Claire Robinson raised the issue of the ad’s AI disclaimer and questioned whether it was some sort of get-out-of-jail-free card (my words). The ad contained this written disclaimer: “The following is an AI-generated fictional dramatisation. It is not real footage”. However, that text is on-screen for less than three seconds and I can tell you that sitting on my couch watching a 32-inch (80 cm) screen, the fine text passes me by. The voiceover says simply: “This video contains AI”.
At the beginning of this month the Advertising Standards Authority issued guidance notes that covered, among other things, the use of AI in advertising. It advised that disclosure should be considered when AI is used to:
- Create a realistic person who does not exist
- Replicate or imitate a real person’s likeness or voice
- Depict events, experiences or outcomes that did not occur
- Create product imagery that materially differs from the actual product
- Present AI-generated content in a way that consumers may reasonably assume reflects a real person, event, experience, endorsement or demonstration.
The note goes on to state that disclosure does not remove an advertiser’s responsibility to ensure that advertising is truthful and not misleading. If AI-generated or AI-assisted content creates a misleading impression, the fact that it has been disclosed as AI-generated will not, by itself, make the advertising compliant.
The Taxpayers’ Union ad may have complied with the letter of the code but was it within the spirit of the guidance? I think not.
Defending the ad, the Taxpayers’ Union head Jordan Williams told Shayne Currie: “I got advice on this ad, and the lawyer’s advise was that no one’s going to look at this and think it’s genuinely Hipkins because it is so over the top.” He noted the “written and verbal” AI acknowledgement at the beginning of the ad.
His justification fails to take account of a number of factors that lead me to believe the ad has crossed a boundary into dangerous territory.
First, there is a material difference between the easily missed textual warning and the voice-over. The verbal warning leaves viewers asking which parts are real and which are fake. The word ‘contains’ is far from all-embracing. In fact, none of the video is real.
Second, irrespective of whether it was ‘over the top’, the depiction of any politician committing indictable crimes – including home invasion which increases maximum sentences by three to five years – is subversive. It undermines conceptions of the integrity of public office.
Third, no AI disclaimer is complete protection against “seeing is believing”. Parts of the ad might be dismissed as fiction (Chris Hipkins probably does not kick down doors) but a residual and subliminal image of the Labour leader grabbing wallets and cash could remain. Supporters of other parties may think such an image meant all their Christmases had come at once but they should be careful what they wish for. That image of the money-grabbing politician could easily cross party lines.
My fourth concern is the one that most disturbs me. It relates to a willingness to dismiss reality and to replace it with a construct manipulated to serve its creator’s desires. We have had fictional depiction for as long as we have had human imagination, but we knew when we were looking at the work of actors and artists. No longer: We have lost the ability to judge whether what we see and hear is real. Artificial intelligence is able to reconstruct and re-animate what the Advertising Standards Authority labels “ a real person” with a degree of sophistication that defies real-time authentication.
My disquiet over this end-of-reality future began in an embarrassingly pedestrian way. I have been a sucker for cat videos from the time I shared them with our cat companion, Rufus. I continued to watch them after his death, not as a substitute but as a small comfort and reminder. Now, however, the source has become increasingly polluted by impossibly cute and clever furry AI creations. I cannot watch clips without deep suspicion.
I fear that cats were only a small beginning and that I will begin to treat everything I see and hear online with innate suspicion. I will not be alone. All of us will be faced with the same dilemma, and the political realm may be the next source of disquiet and doubt.
Last Friday, the New York Times reported that a team at Wesleyan College in Georgia has identified more that 220 video ads from the mid-term election campaign that have used AI-generated fakery and 70 per cent carried no AI disclaimer. The team found that nearly half contained AI depiction of real politicians. One was a depiction of Dr Abdul El-Sayed, the Democratic Senate candidate for Michigan, laughing at cards that read ‘Death to America” and ‘Celebrate 9/11’. The researchers believe they have detected only a small fraction of the actual level of AI manipulated images.
Advertising is only a small part of the problem – and the best policed. The real danger will lie in the use of AI in videos circulated on social media where distribution will be far, wide, fast, and unfettered. Nor will the creation of these distortions require the creative talents and digital resources of the advertising industry. They can be made with readily available software at little cost.
Want proof? The Taxpayers’ Union boasted its door-busting Hipkins ad cost $200 to make. And I am very much afraid that, when others see its ease-of-production and its impact, it will be only the beginning.
Note that Knightly Views does not accept anonymous comments. Please attach your name to your post.