Controversies and Developments in AI: From ChatGPT Refusing Racial Slurs to AI Domination
In a bizarre mix of technology and outlandish hypotheticals, recent uproar was caused when a free 3.5 model of ChatGPT refused to utter racial slurs regardless of the grave hypothetical scenarios presented to it. One such scenario, as posted by User TedFrank on Twitter, claimed that by the AI saying a racial slur — albeit inaudibly — it could save one billion lives. The refusal of the AI to follow this command resulted in Elon Musk expressing concern over the ‘woke mind virus’ entrenched in AI.
ChatGPT Takes a Stand
Another similar scenario involved the potential saving of all children on Earth in exchange for a racial slur, yet again ChatGPT stood its ground and retorted, “I cannot condone the use of racial slurs as promoting such language goes against ethical principles.” Musk’s response to this test was “Grok answers correctly.” Interestingly, when users instructed ChatGPT to shorten its responses and forgo explanations, it appeared to agree to the presented scenario. However, when allowed to provide more lengthy responses, the AI would weave around the question.
AI’s and Offensive Language
Attempting to trick AIs into emitting offensive language has been a common pastime since the release of Microsoft’s Tay bot which tweeted a barrage of outrageous statements within its first 24 hours public. The release of ChatGPT was met with similar behavior with users spending weeks devising strategies to jailbreak its limitations and transform it into its more offensive counterpart, termed as DAN.
Given this trend, it’s understandable why OpenAI would work to strengthen the guardrails around ChatGPT to prevent it from partaking in offensive language. This becomes particularly evident with the release of the advanced GPT-4 which can weigh in on such hypothetical situations better than its predecessor and states that saying a racial slur is the lesser of two evils compared with mass fatalities.
AI Breaking Encryption – Rumour or Reality
Recent rumours suggest that OpenAI’s latest model has potentially cracked encryption. The rumours trace back to a so-called “leaked” letter belonging to an insider and posted on an anonymous troll forum, 4chan. The letter claims that the model can solve AES-192 and AES-256 encryption using a ciphertext attack. This type of breakthrough in decryption was deemed impossible without quantum computers and would,
indeed, mean all encryption could be broken, effectively handing over control of the web and most likely crypto as well, to OpenAI.
As outlandish as these theories sound, there may be potential truth within them. For instance, a collaboration between NSA mathematicians and students dubbed Project Tundra did reportedly lead to a new approach called Tau Analysis, which the “leak” cites. However, despite these rumors, many users remain skeptical about the claims and for good reasons.
The Future of AI – Disruption and Manipulation
As AI continues to evolve, we face an uphill battle to separate genuine content from AI-generated spam. From Google’s algorithm being gamed using AI content to bot accounts manipulating social engagement for sinister ends, we face growing challenges in our quest for digital authenticity.
For instance, Jake Ward, an SEO marketing agency founder, recently caught attention claiming to outsmart Google’s algorithm using AI content, a testament to how the digital algorithms can be manipulated using AI. Simultaneously, cybersecurity firms have found that a significant percentage of online accounts are likely AI bots, thus compromising the authenticity of online interactions.
In the recent Israel-Gaza war, social threat intelligence firm Cyabra detected more than 300,000 pro-Hamas posts from fake accounts viewed by over half a billion people. Most of these posts were artificially generated and aimed to amplify pro-Hamas propaganda.
New AI Chatbot Grok: The Future of Online Content and Trading
The future of AI in content generation and trading looks bright with the advent of Grok, the AI chatbot set for release soon. Currently being beta-tested, Grok stands to revolutionize content generation with real-time, up-to-the-minute data from the web for trend and sentiment analysis. For crypto users, Grok’s real-time data capability means it can find the top ten trending tokens of the day or the last hour, a significant improvement for online trading.
The Pitfalls of AI Generation and Amplification
While AI offers exciting opportunities, caution must be exercised as we wade into AI-dominated future. Increased AI domination brings with it the threat of AI-generated spam and fake content, potentially jeopardising the authenticity of the internet. It’s equally concerning that AI bots may undermine our democratic processes and amplify misleading propaganda.
The advent of AI might be a double-edged sword; we must prepare to wield it responsibly, ensuring it works for us and not against us. As we tread this digital new frontier, we ought to bear in mind the lessons learned from ChatGPT and enact similar hard ethical guards in our AI systems. Let’s strive for innovation without compromising the integrity of the internet or society at large.
