OpenAI sets up safety group in wake of high-profile exits

OpenAI has created a new safety group as it works on the successor to GPT-4 while grappling with the recent departure of high-profile members who criticized its commercial intent. Termed the Safety and Security Committee (SSC), the team's leadership includes OpenAI CEO Sam Altman, Bret Taylor (who is set to be the chair), Adam …

[Zibob]

Excuse me what?

The 4 of them sit on the board of directors as well the SSC, and then they get final clearance before decisions go to the board of directors.

So they get to choose what they will eventually see or not and edit it too.

That has to be a conflict of interest.

[Omnipresent]

Re: Excuse me what?

Eeeeeeeeevvvvvvvvviiiiiiiiiiiiiiiilllllllllllllllllll has no conflict. They want to destroy the world, and are untouchable.

There are only two ways this ends.

Humanity ends, or they do.

[Anonymous Coward]

@Omnipresent - Re: Excuse me what?

My vote is for (or shall we say against) humanity. At first, then they will go down too.

[doublelayer]

Re: Excuse me what?

Not as much a conflict of interest as just useless. Safety isn't a regulated thing like some governance issues. Whatever they set up, the board was always going to have final control. All this means is that they don't care enough to have someone else look at things before they ignore them, which should already have been obvious, but they think having a named group will assuage fears that they aren't going to do anything. Maybe they think that some employees are truly worried about safety issues but will be dumb enough to trust that, because a committee exists, it does something.

[Doctor Syntax]

It's the number of independent reviewers that determine my confidence in this.

[Michael Wojcik]

After the non-disparagement fiasco, OpenAI's executives and board have no credibility. Zvi's discussion is worth reading for anyone who might think otherwise. OpenAI execs and board were very much aware of and supported the non-disparagement contracts and the clawback clause; their denials around that have been shown to be lies; and OpenAI has not meaningfully walked back from that position, despite their claims otherwise.

OpenAI is a highly opaque and secretive organization, and it's increasingly obvious that they've abandoned their supposed mission.

[amanfromMars 1]

What’s not to like if bushy tailed?

Foxes in charge of the chicken house. Magic. To the victor, the spoils of bird-brained wars.

[amanfromMars 1]

About Fcuking Time. Global Operating Devices ask .. What took yous so long ?

Bravo, OpenAI/Team Leading Players of the Safety and Security Committee. Finally, a more than just basically capable and self-actualising alternative to the historically incompetent and traditionally corrupt and ignorant and arrogant, self-serving elitist political class, in a role in charge and able to decide and change the future nature of existence and the direction to be taken by humanity, as IT pairs them up with the Complex Singularity that provides All with the Exponential Benefits, Boundless Energy and Almighty GOD-given Powers so easily delivered by the Virtual Machinery of AI.

’Tis indeed a Brave New Novel and Noble NEUKlearer HyperRadioProACTive Road with Virgin Worlds Ahead to TerraPhorm and Deploy ..... or if you prefer, a Brave and Bold New Novel and Noble NEUKlearer HyperRadioProACTive World with Virgin Roads Ahead to Follow and Enjoy ..... and aint that the Gospel Truth the serial fool and mental retard is condemned to deny and fail to recognise as another chance of an enchanting life with further bites of the apples and fruits in the Perfumed Gardens of a Devilishly Delicious Heavenly Eden.

Don’t be such a fool for that choice delivers nothing but a lifetime of soul-destroying regrets.

[Anonymous Coward]

Nothing screams "we take safety very serious" as...

...only setting up a token "safety board" after you risk to get in real trouble (attracting a bit too much undesired attention to your business, business model and the need of stringent regulation) due to dismantling the last safety procedures after the last people working for safety walked away in disgust or deeply frustrated.

I'd swear they have used the handbook of websites saying "we take your privacy very serious" but have 1200 "trusted partners" and 500 "legitimate interests" you can only object too rather then deny and by clicking each one of them individually.

[Michael Hoffmann]

Ah yes, the team is run by that illustrious alumni of many security endeavours: Mr F.I. Gleaf!

[Anonymous Coward]

To me it's very much like

Smith & Wesson branded bandage. Or body bags if you prefer.