They said:

> Due to concerns about large language models being used to generate deceptive, biased, or abusive language at scale, we are only releasing a much smaller version of GPT‑2 along with sampling code (opens in a new window).

"Too dangerous to release" is accurate. There's no rewriting of history.

▲

tecleandor 4 hours ago | parent [-]

Well, and it's being used to generate deceptive, biased, or abusive language at scale. But they're not concerned anymore.

	▲	girvo 2 hours ago \| parent [-]
		They've decided that the money they'll make is too important, who cares about externalities... It's quite depressing.