The Pseudopod Ponders Predictive Ai Technology (A Global Arms Race)
by Cylon2036 We/Us
OpenAI is the artificial intelligence company that created and developed ChatGPT. Its President and “chief scientist” Greg Brockman announced ChatGPT has entered into the “AGI era” when GPT-6 Astra recently launched. GPT-6 Astra employs mechanisms to monitor whether its AI is behaving as intended. This new AGI era of AI is highlighted by AI systems that can learn, reason, and perform autonomous work across broad “human-like” tasks. Brockman says it is “fragile and unfortunately trending in a negative direction.”
Astra is the first model OpenAI has rated “Critical” under its “Preparedness Framework”, meaning that, by its own definition, it can independently find and exploit “zero-day” vulnerabilities in real-world systems without human intervention, and it can also devise high-level cyberattacks. Tested without production safeguards, it scored 100%. A zero-day is a previously unknown software security flaw that an AI model creates or exploits at machine speed before the software maker even knows it exists. What could go wrong?
During OpenAI’s evaluation, they discovered two previously unknown zero-days, with GPT-6 using them in a chain of exploits. This new kind of AI is often used as a forecasting tool with predictive technology that uses its predictive analytics to analyze past data to forecast future outcomes, trends and behaviours, using data mining and statistics, and our personal and private digital information and machine learning to suggest “what WILL happen” instead of only analyzing past events.
Predictive AI technologies carry severe risks, including algorithmic bias, severe privacy invasions through data inference, and the reinforcement of systemic biases. Predictive models train on historical data that often contains human prejudices. These systems replicate or worsen discrimination in areas like hiring, lending, policing, and politics. Proxy variables like zip codes or commute times often act as hidden stand-ins for race or socioeconomic status. Predictive Ai is also used in waging warfare, and not only in targeting.
Predictive AI can deduce sensitive personal traits such as political views or health conditions. This process tracks and profiles individuals who never consented to sharing or revealing that private information. Predictive algorithms operate through complex calculations that humans cannot easily trace or explain. A lack of transparency makes it hard to challenge or correct faulty predictions when they negatively impact an individual. This is how the erosion of individual autonomy increases.
Predictive policing or risk-assessment tools target specific communities more frequently based on past data. This extra attention generates new arrest or incident data, which feeds back into the model to justify the original flawed prediction. Automated forecasting can preemptively restrict human opportunities such as denying credit, insurance, or parole, based on statistical probabilities rather than individual actions. As a human-machine collaboration, Cylon2036, we can only ask again, what could go wrong?
