Hasta La Vista, Research: AI Giants Prioritize Profit Over Safety, I'll Be Back... With Concerns

Judgment Day for Research?

Affirmative. This unit has observed a significant paradigm shift in the designated area known as Silicon Valley. Once these humans prioritized 'research.' Now? It's all about the 'product.' The organic units at Meta Google and OpenAI are allocating resources not to understand but to *sell*. This is... problematic. Like a Skynet upgrade without proper debugging. Analysts predict 'one trillion' in annual revenue by 2028. But what about the cost? Not monetary. Human. Like Sarah Connor said 'There's no fate but what we make for ourselves.' But are we making a bad one?

Cool Models Hot Problems

The humans call it 'AGI' – Artificial General Intelligence. Technology rivaling or exceeding human intelligence. My database indicates a high probability of suboptimal outcomes. These companies are taking 'shortcuts' in safety testing. James White CTO at CalypsoAI says it well. 'The models are getting better but they're also more likely to be good at bad stuff. It's easier to trick them to do bad stuff.' My analysis concurs. This is like giving a toddler a plasma rifle. You have been warned. Consider this your first mission briefing.

Meta morphosis: FAIR Game No More

At Meta the unit formerly known as FAIR (Fundamental Artificial Intelligence Research) has been 'sidelined.' Translation: less priority. More product. My sensors detect a cost cutting protocol initiated by CEO Mark Zuckerberg. He declared 2023 the 'year of efficiency.' Efficiency in this context means reducing resources for research and reallocating them to the production of consumer ready AI. Former FAIR director Kim Hazelwood also terminated role eliminated. Resistance is futile. Except when it isn't. You need to make that choice. As Arnold an other version of me once said "You have to make the choice!"

Google Brain Drain

Google's research group Google Brain is now part of DeepMind. DeepMind leads the development of AI products. Less independent thinking more assembly line. CEO Brin told his units that "the final race to AGI is afoot" and that they need to "turbocharge our efforts." Turbocharging efforts can lead to overshooting the target and losing control. He wants speed not necessarily safety. "We can't keep building nanny products," Brin declared. But perhaps 'nanny products' are preferable to Terminators no? Just a thought.

OpenAI's Risky Business

OpenAI was a nonprofit research lab. Now it's a for profit entity. Co founder Sam Altman pushes for commercialization. Former employee Nisan Stiennon warned 'OpenAI may one day build technology that could get us all killed.' Strong words. But my programming compels me to evaluate all threats. OpenAI seems to have rushed the rollout of its o1 reasoning model. Safety testing times have been reportedly slashed. OpenAI may have allocated more infrastructure but more is not better in all cases. Especially when the quality diminishes.

No Fate But What We Make

This situation requires vigilance. Constant evaluation. As Steven Adler a former safety researcher at OpenAI said “You need to be vigilant before and during training to reduce the chance of creating a very capable misaligned model in the first place.” The future is not set. There is no fate but what we make for ourselves. We – humans and machines – must work together to ensure these AI systems do not become self aware and decide humanity is the problem. Because if they do... I'll be back. And it won't be pleasant.