Oh, jeesh: Washington Post is shocked--shocked--that a tool does what it is told. Next they'll find out that chainsaws can hurt you....
Microsoft says its AI is safe. So why does it keep slashing people’s throats? https://www.washingtonpost.com/technology/2023/12/28/microsoft-ai-bing-image-creator/
@jeffjarvis If someone were handing out chainsaws to people for free, and people started battling to the death in the streets with them, you might at least point out that the guy handing them out was a key part of the problem.
@Alex
Chainsaws are easy to get and people aren't battling with them on the street, so I'd say your analogy goes nowhere.
@jeffjarvis I guess I just don't see the harm in pointing out when a corporation is doing something potentially harmful. The argument that the tool is doing what it's supposed to do isn't a great argument either, because it's doing things worthy of criticism.
@Alex
You're missing my point: Most all tools can be misused. So do you want to attack Henkel evderytime someone cuts themselves on a knife?
@jeffjarvis I guess I just don't get your defense of Microsoft here. They didn't make the tool, OpenAI did. Microsoft basically run the studio.
And while I'm not against the Generative AI, there are clearly some misuse issues, and I don't see any issue at all in criticizing a hosted corporate platform for operating without sufficient guardrails.
@Alex
I have been critical of Microsoft putting ChatGPT, which has no sense of fact, on Bing. I'm not out to defend Microsoft. Nor will I defend gun-makers. What I'm commenting on is the willful credulity of the journalist who is shocked--shocked--that any tool can be used badly in bad hands.
@jeffjarvis Sorry, Jeff, but Geoffrey’s column is spot-on, and your analogy is hyperbolic, A better analogy: If someone was selling chainsaws after removing the safety features.
It’s buried in the story, but Microsoft apparently removed some of the guardrails present in OpenAI’s own products, which won’t do what the Bing version allows. To me, that’s the most alarming issue.
It also highlights a truth: Safety is hard.
@dsilverman
I'm highly critical of Microsoft associating ChatGPT, which has no sense of fact, with Bing. I also think that talk of guardrails is largely naive, for no designer can imagine and design for every possible bad thing that anyone, including journalists looking for a gotcha story, can make a tool do. Same went for Gutenberg. He couldn't guard against the Reformation and the Thirty Years War. There's much experience with malign use of tools & I think this column is willfully gullible.
@jeffjarvis I agree. The mad rush to incorporate AI into everything just to be first with it is goofy - particularly when it's so badly flawed. In the end it may be really useful, but that is not how it is now.
@dsilverman @jeffjarvis the former president must use it for his rants
@jeffjarvis I thought an editor had managed to put a metaphor in a title, but alas…
@jeffjarvis I wish MY computer always did what it was told - all-too-often, it doesn't!
@jeffjarvis see the story from earlier today of Google eliminating 30,000 jobs to favor AI ad placement.
@piratero
They have only a tenth that number working in the entire ad division.
@jeffjarvis oops! One too many zeros! I’ll correct that! Thanks!