Now, only damage control is possible

…With Mythos, the watch­man who is sup­posed to identify the vul­ner­ab­il­it­ies and cracks is also the only one who can pro­tect the sys­tem by tak­ing an eth­ical pos­i­tion. Except Mythos has no sense of eth­ics. Anthropic might loosely have one, but not enough to stop it from releas­ing the model, even in lim­ited pre­view.

Tech­no­lo­gies are not tools. they are pro­pos­i­tions. They ima­gine the worlds that cluster around them, and then begin to make those worlds avail­able, prof­it­able, and inev­it­able. They are not dan­ger­ous only when deployed. They can be dan­ger­ous in their very exist­ence. We now have the nuc­lear ver­sion of gen­er­at­ive AI sit­ting with us. Not because it will des­troy the world tomor­row, but because its exist­ence reor­gan­ises the con­di­tions of secur­ity, vul­ner­ab­il­ity, trust, and power. Anthropic might stall this for a while, but oth­ers will get there and not per­form this lim­ited restraint. Once a capa­city exists, it becomes nor­m­al­ised, and we adapt to it.

Full article: https://reurl.cc/N2Oxpk (Indian Express)