RHETORWRECK

Rhetoric taken apart.
Politics, media and AI. Naming the moves that hide responsibility and choice.
Ali Graham

[RHETORWRECK @Rhetorwreck]

The word intelligence makes a claim that performance alone does not settle.

What does the evidence actually support?

The Rhetorwreck of AI: Named for intelligence it has yet to demonstrate.

[RHETORWRECK @Rhetorwreck]

@Paul__Walsh “Gone rogue” works in the same way.
Corporate responsibility is shielded.
See my article:

[RHETORWRECK @Rhetorwreck]

Anil Seth’s biological naturalism is strongest when “being alive” is an empirical claim about what consciousness requires.

It becomes circular if “consciousness” is defined in a way that simply excludes the non-biological.

Other qualities of intelligence are reviewed in my article:

[RHETORWRECK @Rhetorwreck]

Anil Seth’s biological naturalism is strongest when “being alive” is an empirical claim about what consciousness requires.

It becomes circular if “consciousness” is defined in a way that simply excludes the non-biological.

Other qualities of intelligence are reviewed in my article:

[RHETORWRECK @Rhetorwreck]

When an AI system does something its creators did not prevent, we’re encouraged to say it has “gone rogue”.

I wrote about what that phrase does to responsibility, see please my article.

[RHETORWRECK @Rhetorwreck]

“Rogue” is becoming the default word for behaviour produced by systems companies built, trained and released.

Notice what happens to responsibility when the machine becomes the rogue.

Who went rogue?

[RHETORWRECK @Rhetorwreck]

@FT AI tools look impressive in a paper, benchmark or demo. Its efficacy is only established in use: across real institutions, real workflows and real people, over time.
Note these real world case studies often go unmentioned in performance claims.

[RHETORWRECK @Rhetorwreck]

@notYurpi @alex_verem This touches on something I’ve been writing about: by calling these systems “AI” we get ahead of what the technology demonstrates.
See my article,

[RHETORWRECK @Rhetorwreck]

Once, we gave a search engine a butler because we needed a human metaphor for a new machine.

Now we put intelligence in the name.

Is the name “AI” the Ask Jeeves of our time? https://t.co/2KLOwKdvBa

[RHETORWRECK @Rhetorwreck]

Yes. This is exactly the distinction I was trying to get at in my article on AI “going rogue”.

If a model is instructed to conduct a cyberattack, then it does so in an environment whose boundaries were left open.

The words move responsibility from the people who designed the task to the system that carried it out.

[RHETORWRECK @Rhetorwreck]

@FT “AI” is becoming a political container.

For Sanders: corporate power and jobs. For Bannon: sovereignty and social disruption. For others: surveillance, environmental costs of data centres, extinction.

The label hides the people making the choices. Not for much longer.

[RHETORWRECK @Rhetorwreck]

@FT “AI” is becoming a political container.

For Sanders: corporate power and jobs. For Bannon: sovereignty and social disruption. For others: surveillance, environmental costs of data centres, extinction.

The label hides the people making the choices. Not for much longer.

[RHETORWRECK @Rhetorwreck]

Carl Sagan’s warning is timely.
Most of us don’t understand the technology.
We’re also encouraged to make claims about it in the words used to describe its performance.
Intelligence, reasoning, thinking, agency, consciousness, even play.
The vocabulary closes the gap before the science has been demonstrated.

[RHETORWRECK @Rhetorwreck]

How does someone who once had an extraordinary instinct for an audience begin to lose it?

By continuing to sound exactly like the person who won it in the first place.

The Voice Trap. https://t.co/Kr4PxDNBSw

[RHETORWRECK @Rhetorwreck]

@IntCyberDigest Bjarne Stroustrup’s point cuts through the rhetoric on “AI coding”.

Generating code is not the same as understanding what the code needs to do.

Whether it is safe, &c.: judgement stuff.

[RHETORWRECK @Rhetorwreck]

@IntCyberDigest Bjarne Stroustrup’s point cuts through the rhetoric on “AI coding”.

Generating code is not the same as understanding what the code needs to do.

Whether it is safe, &c.: judgement stuff.

[RHETORWRECK @Rhetorwreck]

@thesupermanmx We named the field too soon.

If language models aren’t “general intelligence”, adding vision doesn’t make the name any less demonstrated.

Ask Jeeves wasn’t a butler. “Visual General Intelligence” isn’t sufficient evidence of general intelligence.

[RHETORWRECK @Rhetorwreck]

@thesupermanmx We named the field too soon.

If language models aren’t “general intelligence”, adding vision doesn’t make the name any less demonstrated.

Ask Jeeves wasn’t a butler. “Visual General Intelligence” isn’t sufficient evidence of general intelligence.

[RHETORWRECK @Rhetorwreck]

@FT AI may have a PR problem. It also has a language problem.

The industry describes its choices as inevitabilities: intelligence will scale, jobs will change, data centres are required, transformation is coming.

Trust is hard to rebuild when responsibility is left out.

[RHETORWRECK @Rhetorwreck]

@FT AI may have a PR problem. It also has a language problem.

The industry describes its choices as inevitabilities: intelligence will scale, jobs will change, data centres are required, transformation is coming.

Trust is hard to rebuild when responsibility is left out.