The Unreal #TechVision Now if it sounds like Hour One is creating just synthetic AI but AI in general has played a bots, and bots accounted for 38% of tweets 140 deepfakes – or fake videos of real people saying major role in accelerating this sense of mistrust. about fake science. Moreover, research or doing things they haven’t done – that’s Some believe the use of AI algorithms in social from Carnegie Mellon University’s Center for because it is. But given the videos are safe and media to personalize content has led to filter Informed Democracy and Cybersecurity found legal, the actors have given permission for the bubbles, echo chambers, and algorithmic that bots may account for 45% to 60% of 141 use of their likeness, and the videos are clearly confounding, leading users to be shown more Twitter accounts discussing COVID-19. 136 labeled as computer-generated, Hour One limited and extreme content. Additionally, the seems far from malicious. So does that mean methods by which social media companies’ Enter the advancing capabilities of generative it’s okay? This points to the potential challenges algorithms prioritize engagement can lead to AI coupled with social media algorithms that and pitfalls that can arise with the use of unintended outcomes. False news online tends prioritize engagement, and it’s no wonder we’re synthetic data – and so far, its use has not to diffuse faster and further than the truth – in an infodemic with lowered trust. And as AI 137,138 progresses and we enter a world of synthetic been entirely good. Inadvertently, synthetic making the truth very much at risk. data has left out important outliers, failed to realness, things could get much worse. accurately increase diversity, and amplified To illustrate, it was revealed that prior to the bias too. And purposely malicious deepfakes 2020 U.S. election, Eastern European troll farms and disinformation have caused damage. were able to reach 140 million people a month on social media with provocative propaganda, All this comes at a time when trust in the 75% of whom saw the content because of 139 technology sector is on a steep decline, reaching the recommendation-engine. The use of AI 135 all-time lows in 17 out of 27 countries. Trust in social media bots can also amplify stories. in information sources across all media is at a Research from Brown University found that record low too, with social media in last place on an average day during their study, 25% of (35% of those surveyed trust social media). Not tweets concerning the climate crisis were from Introduction // WebMe // Programmable World // The Unreal // Computing the Impossible 63
Report Page 62 Page 64