Have I hardened against LLMs? (www.baldurbjarnason.com)

In a recent reflection, a tech writer revisited their stance on large language models (LLMs) and other generative technologies. While their technical views on LLMs remain largely unchanged since the publication of their book—highlighting significant downsides and narrow use cases—their perspective on the tech industry has shifted dramatically. They expressed disillusionment with how industry leaders and developers continue to overlook the ethical implications and potential harms of these technologies, such as the propagation of child sexual abuse material (CSAM), misinformation, and insecure software practices, in favor of marginal productivity gains. The writer criticizes the industry for disregarding substantial risks associated with LLMs and for enabling harmful behavior through these platforms. They note a troubling trend where discussions around LLMs focus on potential benefits, often dismissing ethical concerns raised by experts as overly cautious or extremist. The writer concludes that, despite the ongoing flaws in LLM technologies, the industry's response to these issues has been disappointing, highlighting a worrying prioritization of profit over responsibility. The only exception they find valuable is in certain speech recognition tools, which they argue are the sole beneficial outcome from the current generative model landscape, barring OpenAI's offerings that are marred by inaccuracies.