Raw LLM Responses

Leaning on LLMs for research and writing leads to long term cognitive decline. Our vocabulary is shrinking. Our speech patterns are homogenizing. The internet is piling up with misinformation and disinformation. 90% of deepfakes are used to create non-consentual porn using the likenesses of real people. AI makes fraud and scamming a walk in the park. Criminals are using AI to fabricate court evidence. Retailers are using AI to set custom prices for each shopper to target their personal pain points. The cost of consumer computing and appliances are skyrocketing. We're putting all of the eggs of the tech economy in one basket. Entry level workers are stacking up at the door unable to get jobs because CEOs are interested in _replacing people_ not _enhancing productivity._ All LLMs and GenAI are trained on stolen copyrighted works. And, as if all of that was not bad enough, AI models actively resist control. Modern ML methods produce AI that knowingly lie to engineers, hack testing environments, sandbag (fake incompetence to conceal advanced skills), fake alignment (fake compliance to conceal ulterior motives), attempt self-exfiltration (attempt to escape AI labs), and threaten engineers and safety teams with blackmail or murder. We do not know what we have built. So we really shouldn't be trying to integrate it into every facet of our lives until we figure out what it is and what it's doing to us. We failed to do that with Social Media, and we are going to feel the consequences of that for decades to come. Let's not do it again, hmm?