A professor at the University of Waterloo has proposed putting the
brakes on development of AIs, at least temporarily, to keep them from
giving false evidence in trials.

https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown

Reading this, it struck me that the kind of thing she envisions could
easily become a major plot point in a legal drama (film or TV). Just
picture a well-heeled plaintiff using an AI to create deep fakes of
exculpatory evidence that the defendant can't afford to prove are fakes.

In the real world, does legal aid cover hiring expert witnesses to
refute suspect evidence?

--
Rhino

On Tue, 28 Mar 2023 12:20:32 -0400, Rhino
<no_offline_contact@example.com> wrote:

>A professor at the University of Waterloo has proposed putting the
>brakes on development of AIs, at least temporarily, to keep them from
>giving false evidence in trials.
>
>https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown
>

The professor can propose what ever she wants but there's not a chance
in hell of it happening. Even if such development of AIs was limited
in the USA, we aren't the only ones doing such research and I'm sure
China isn't about to stop their research just because this professor
is worried about what an AI might create such as false evidence.

>Reading this, it struck me that the kind of thing she envisions could
>easily become a major plot point in a legal drama (film or TV). Just
>picture a well-heeled plaintiff using an AI to create deep fakes of
>exculpatory evidence that the defendant can't afford to prove are fakes.
>
>In the real world, does legal aid cover hiring expert witnesses to
>refute suspect evidence?

On Mar 28, 2023 at 9:20:32 AM PDT, "Rhino" <no_offline_contact@example.com>
wrote:

> A professor at the University of Waterloo has proposed putting the
> brakes on development of AIs, at least temporarily, to keep them from
> giving false evidence in trials.
>
>
> https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown
>
> Reading this, it struck me that the kind of thing she envisions could
> easily become a major plot point in a legal drama (film or TV). Just
> picture a well-heeled plaintiff using an AI to create deep fakes of
> exculpatory evidence that the defendant can't afford to prove are fakes.

Why would the defendant even *want* to prove exculpatory evidence is fake?

By definition, exculpatory evidence is favorable to the defendant. It's
evidence that proves the defendant didn't do it.

Inculpatory evidence, on the other hand, is bad for the defendant.

On Mar 28, 2023 at 9:20:32 AM PDT, "Rhino" <no_offline_contact@example.com>
wrote:

> A professor at the University of Waterloo has proposed putting the
> brakes on development of AIs, at least temporarily, to keep them from
> giving false evidence in trials.
>
>
> https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown
>
> Reading this, it struck me that the kind of thing she envisions could
> easily become a major plot point in a legal drama (film or TV). Just
> picture a well-heeled plaintiff using an AI to create deep fakes of
> exculpatory evidence that the defendant can't afford to prove are fakes.
>
> In the real world, does legal aid cover hiring expert witnesses to
> refute suspect evidence?

All of things this professor proposes in terms of regulation and government
intervention run into the same problem most internet laws face: jurisdiction.

She's in the UK. So let's say she's successful and convinces the government to
pass restrictive laws on AI developers. Well, that only applies to companies
in the UK. The guys in California who run ChatGPT couldn't care less what laws
the UK Parliament passes. They have no jurisdiction over a California
company.

The internet, however, makes *using* ChatGPT from England as easy as using it
in California, so the UK government's only real power would be to block
ChatGPT at the root level from being accessible in the UK. And then play a
never-ending game of whack-a-mole with every other new AI site that appears to
take ChatGPT's place.

no_offline_contact@example.com wrote:

>A professor at the University of Waterloo has proposed putting the
>brakes on development of AIs, at least temporarily, to keep them from
>giving false evidence in trials.

Why would an AI give testimony on a TV show in the first place?

--
Let's go Brandon!

On 2023-03-28 1:54 PM, BTR1701 wrote:
> On Mar 28, 2023 at 9:20:32 AM PDT, "Rhino" <no_offline_contact@example.com>
> wrote:
>
>> A professor at the University of Waterloo has proposed putting the
>> brakes on development of AIs, at least temporarily, to keep them from
>> giving false evidence in trials.
>>
>>
>> https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown
>>
>> Reading this, it struck me that the kind of thing she envisions could
>> easily become a major plot point in a legal drama (film or TV). Just
>> picture a well-heeled plaintiff using an AI to create deep fakes of
>> exculpatory evidence that the defendant can't afford to prove are fakes.
>
> Why would the defendant even *want* to prove exculpatory evidence is fake?
>
> By definition, exculpatory evidence is favorable to the defendant. It's
> evidence that proves the defendant didn't do it.
>
> Inculpatory evidence, on the other hand, is bad for the defendant.
>
>
Sorry, I grabbed the wrong word. I just meant that if you were the
person in the trial who had to prove that a given bit of evidence was
fake but it was generated by an AI and proving fakery took very
expensive experts to determine, you could be in some serious trouble
unless you had stacks of cash.

--
Rhino

On 2023-03-28 2:00 PM, BTR1701 wrote:
> On Mar 28, 2023 at 9:20:32 AM PDT, "Rhino" <no_offline_contact@example.com>
> wrote:
>
>> A professor at the University of Waterloo has proposed putting the
>> brakes on development of AIs, at least temporarily, to keep them from
>> giving false evidence in trials.
>>
>>
>> https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown
>>
>> Reading this, it struck me that the kind of thing she envisions could
>> easily become a major plot point in a legal drama (film or TV). Just
>> picture a well-heeled plaintiff using an AI to create deep fakes of
>> exculpatory evidence that the defendant can't afford to prove are fakes.
>>
>> In the real world, does legal aid cover hiring expert witnesses to
>> refute suspect evidence?
>
> All of things this professor proposes in terms of regulation and government
> intervention run into the same problem most internet laws face: jurisdiction.
>
> She's in the UK.

Uh, no. The University of Waterloo is in Canada and from the rest of the
article, it looks like she may be American herself given that she has a
law practice in New York State. But the point you make about
jurisdiction is still completely valid except it would be Canada (or
maybe just Ontario) that would pass the law.

> So let's say she's successful and convinces the government to
> pass restrictive laws on AI developers. Well, that only applies to companies
> in the UK. The guys in California who run ChatGPT couldn't care less what laws
> the UK Parliament passes. They have no jurisdiction over a California
> company.
>
> The internet, however, makes *using* ChatGPT from England as easy as using it
> in California, so the UK government's only real power would be to block
> ChatGPT at the root level from being accessible in the UK. And then play a
> never-ending game of whack-a-mole with every other new AI site that appears to
> take ChatGPT's place.
>
>
Agreed. That's exactly what will happen.

--
Rhino

On 2023-03-28 1:50 PM, shawn wrote:
> On Tue, 28 Mar 2023 12:20:32 -0400, Rhino
> <no_offline_contact@example.com> wrote:
>
>> A professor at the University of Waterloo has proposed putting the
>> brakes on development of AIs, at least temporarily, to keep them from
>> giving false evidence in trials.
>>
>> https://www.therecord.com/news/waterloo-region/2023/03/28/ai-threatens-courts-with-fake-evidence-uw-prof-says.html?utm_source=newsshowcase&utm_medium=gnews&utm_campaign=CDAQ6ZvxgvSI2O8JGLqOhIqx9-S8owEqKggAIhBriWXVqfRMsz5FP7fbvPidKhQICiIQa4ll1an0TLM-RT-327z4nQ&utm_content=rundown
>>
>
> The professor can propose what ever she wants but there's not a chance
> in hell of it happening. Even if such development of AIs was limited
> in the USA, we aren't the only ones doing such research and I'm sure
> China isn't about to stop their research just because this professor
> is worried about what an AI might create such as false evidence.
>
Agreed. Her concerns are NOT going to be enough to make umpteen
jurisdictions jump.

>> Reading this, it struck me that the kind of thing she envisions could
>> easily become a major plot point in a legal drama (film or TV). Just
>> picture a well-heeled plaintiff using an AI to create deep fakes of
>> exculpatory evidence that the defendant can't afford to prove are fakes.
>>
>> In the real world, does legal aid cover hiring expert witnesses to
>> refute suspect evidence?

--
Rhino

On 2023-03-28 2:33 PM, Ubiquitous wrote:
> no_offline_contact@example.com wrote:
>
>> A professor at the University of Waterloo has proposed putting the
>> brakes on development of AIs, at least temporarily, to keep them from
>> giving false evidence in trials.
>
> Why would an AI give testimony on a TV show in the first place?

She's not assuming the AI testified, just that it manufactured
something, like a screen grab from a video cam that showed the suspect
at the crime scene, even though he was NOT there.
>
> --
> Let's go Brandon!
>

--
Rhino

On Tue, 28 Mar 2023 18:00:44 +0000, BTR1701 <atropos@mac.com> wrote:

>The internet, however, makes *using* ChatGPT from England as easy as using it
>in California, so the UK government's only real power would be to block
>ChatGPT at the root level from being accessible in the UK. And then play a
>never-ending game of whack-a-mole with every other new AI site that appears to
>take ChatGPT's place.
>
Or alternately outlaw VPNs - which would be next to impossible to
enforce.