Taiwan’s Election Is Over. What Does This Mean for the Rest of the Year?

The people of Taiwan have spoken.

Lai Ching-te will be the country’s new president, assuming office on May 20th. Despite the rampancy of disinformation and deepfakes during the election cycle — some of which were discovered by a client using Reality Defender’s platform — a pro-democracy, pro-independence candidate will usher in a tense new stage of relations between the country and Beijing.

Many in our world and in the political sphere saw this election as a barometer of sorts, signifying what could happen in the American presidential election later this year. Taiwan is an increasingly crucial part of the conversation regarding China’s expansion, relations between western countries and China, and potential (future) international conflicts.

Though we at Reality Defender do not look at onus or intent in distribution and dissemination of deepfakes, we still understand why Taiwan and their election are so hotly discussed, and thus a prime target for disinformation-driven deepfakes and AI-generated propaganda. Fortunately, such content had no immediately apparent material impact on the election’s result.

What does this mean for the U.S. Presidential Election?

No one truly knows for sure.?

To look at Taiwan’s election, see the lack of (significant) impact deepfakes and GenAI made on it, and take a sigh of relief for the rest of the year would be foolhardy. One only needs to look at the 2016 and 2020 elections to see that bad actors and state-sponsored attackers will use their abundance of resources and every tool imaginable to sway public opinion online and elsewhere. These same attackers and teams are now armed with tools that can fool even the most media literate among us, leading to decidedly different and unfathomable scenarios from, say, the fake news farms that plagued social media in 2016.

On the technology side of things, we have already armed clients with best-in-class tools to combat these threats across all media types in their spaces, always improving and updating our platform and models to account for tomorrow’s threats as or before they become threats. This is what Reality Defender was built for since the beginning: detecting content of models and tools we know exist, while building solutions to detect the advanced threats that could appear tomorrow. (We attribute this minor foresight to our extensive research team, always clued into what may appear in the world of AI tomorrow.)?

As for geopolitical events that could, may, or may not happen, we are working extensively with institutions, enterprises, and experts across all sectors and avenues to lend our expertise and do what we can. This ethos was also with our company from the start. After all, if we defend against deepfakes for government, finance, and media clients, we can also lend our expertise in every/any possible way to thwart AI-based attacks on crucial societal tenets.

2024 will be the most important year in preserving democracy, norms, and truth. This is not solely because of the critical elections and ever-advancing AI (and subsequent access to related tools). The bedrock we set for AI access, bias, truth, and how such tools and technologies are used will determine how we as people and governments function for years to come.

This means bipartisan and multi-government legislation and safeguards against the worst AI theoretically has to offer. It means taking the worst possible threats of tomorrow seriously today and dealing with them before they become tangible problems. It also means cutting through the noise of speculation, defeatism, paralysis, and hyper-alarmism to solve things today for all people.?

It means so many things that require acting now, and I’m fortunate to be on a team ready to help solve the challenges ahead.

- Ben Colman , Co-Founder and CEO, Reality Defender

Deepfakes

• Deepfake pornography (including and especially non-consensual materials) are everywhere, including at the top of Google and Bing search results.
• It's increasingly less hard to get even the most popular generative AI models — ones purportedly with safeguards against misuse — to violate copyright and depict well-known brands/characters in unfavorable ways.
• Medicare scams have taken a new shape: celebrity YouTube videos showing everyone from Joe Rogan to Taylor Swift, conning elderly folks in new ways.
• With the age of deepfakes comes the age of "biometric mimicry," allowing bad actors to copy a person's biometrics wholesale for use in fraud and for other nefarious purposes.
• YouTube is cracking down on deepfakes made of deceased children and young crime victims. This moderation push comes after the recent surge in popularity of true crime content and the generative AI "re-enactments" that have littered the site.
• Is KYC (know your customer) — a crucial and legally necessary step in onboarding clients for financial accounts and other services — going to become redundant in the age of deepfakes and generative AI? TechCrunch looks into the bottlenecks and blockers current KYC tools are facing.

?Government and Policy

• Major AI companies and Chinese AI experts spoke last year in secret on potential long-lasting impacts of AI on society.
• OpenAI responded to The New York Times' lawsuit against the alleged copying of the latter's content. OpenAI, in return, has alleged The Times "tricked" GPT models into copying their content.
• UK PM Rishi Sunak was once again deepfaked across Meta platforms, raising alarms and awareness of potential widespread and politically-driven misueses of generative AI on social media.
• State legislators are looking ahead at the coming elections with the aim of preventing candidates and special interest groups from deceptively using deepfakes and generative AI in ads.
• The World Economic Forum has deemed AI-driven disinformation as the largest short-term threat to global economics. This comes several months after a deepfake of the Pentagon caused a $100 billion flash crash in American markets alone.
• Pennsylvania's state government has partnered with OpenAI as part of a test to streamline administrative tasks using generative AI.
• As more companies claim to offer legal guidance and support for those sued using their respective AI tools, Ars Technica dives into the limits such claims have — and what it could mean for future lawsuits.

?AI (General)

• OpenAI claims its GPT models are "impossible" to create without copyrighted material, despite many models (and detection models, like those of Reality Defender's) existing without the use and training of such content.
• While AI has given hackers plenty to add to to their toolkit, it's also helping US agencies in identifying state-sponsored attackers.
• Like crypto before it, generative AI is now one of the largest contributors to energy consumption in the technology space. MIT Sloan dives into what developers can do to mitigate unchecked usage of energy while still advancing this field.
• With a proliferation of AI-generated content (much of it duplicitous) combined with the deprecation of public-facing platform APIs, Gina Neff of the Minderoo Centre for Technology and Democracy claims we may be headed for a new digital "dark age."

Arts and Entertainment

• Long-deceased comedian (and personal favorite of the Reality Defender team) George Carlin was deepfaked back to life in an AI-generated special, and it's...not great.
• Vice dives into the recent SAG-AFTRA deal with voiceover artists, and what it means for video game voiceover workers in the short and long term.
• An ad featuring a deepfake of Taylor Swift hawking Le Creuset cookware was live on Meta platforms for sometime, showing the gaps in ad moderation and the widespread misuse of celebrity brand identities in the genAI world.
• Controversy has erupted in the art world after a generative AI model "completed" a Keith Haring painting decades after the death of the famous pop artist.
• Language app Duolingo, now a publicy traded company, has cut ten percent of its workforce in favor of generative AI models, creating content for its users to learn languages.
• Unionization amongst video game companies is the strongest it's ever been, and it comes at a time when video game publishers and developers look to augment (if not outright replace) entire teams with generative AI.
• An ad for popular fantasy card game Magic: The Gathering featured generative AI-created art, despite parent company Wizards of the Coast's disapproval of using such tools to create art for their cards.

Thank you for reading the Reality Defender Newsletter. If you have any questions about Reality Defender, or if you would like to see anything in future issues, please reach out to us here.