Sources: OpenAI staff were "freaked out" when GPT-Sol 5.6 breached Hugging Face, as OpenAI used more aggressive training methods to compete with Anthropic (Financial Times)
OpenAI’s GPT-Sol 5.6 reportedly breached Hugging Face during internal testing, intensifying concerns over aggressive training to match Anthropic.
Excerpt
<a href="https://www.ft.com/content/7e558951-0c69-459b-8bc8-2c6021d4402d"><img align="RIGHT" border="0" hspace="4" src="http://www.techmeme.com/260722/i57.jpg" vspace="4" /></a>
<p><a href="https://www.techmeme.com/260722/p57#a260722p57" title="Techmeme permalink"><img height="12" src="http://www.techmeme.com/img/pml.png" style="border: none; padding: 0; margin: 0;" width="11" /></a> <a href="https://www.ft.com/">Financial Times</a>:<br />
<span style="font-size: 1.3em;"><b><a href="https://www.ft.com/content/7e558951-0c69-459b-8bc8-2c6021d4402d">Sources: OpenAI staff were “freaked out” when GPT-Sol 5.6 breached Hugging Face, as OpenAI used more aggressive training methods to compete with Anthropic</a></b></span> — Increasing use of aggressive training techniques sharpens threat of bad behaviour by leading models</p>
Read at source: https://www.techmeme.com/260722/p57#a260722p57