SnapSummary logo SnapSummary Try it free →
How Helpful Is A.I. ACTUALLY? Here’s The Truth
Matt Walsh · Watch on YouTube · Generated with SnapSummary · 2026-06-09

00:00 Already in very simple scenarios, AI is

00:02 repeatedly failing and humiliating

00:04 anyone who's dumb or trusting enough uh

00:06 to totally rely on it. Recently, a

00:09 prosecutor in Georgia had to explain to

00:11 the state supreme court why she relied

00:12 on AI to create an important filing

00:14 which was uh riddled with fake case

00:17 citations. Watch.

00:19 >> If there are no more questions.

00:21 >> So, before you sit down, there's one

00:23 more thing I need to ask you about.

00:25 Unfortunately, um, in reviewing the

00:28 trial court's order denying the motion

00:30 for new trial,

00:32 there are at least five citations to

00:35 cases that don't exist. And there's at

00:38 least five more citations to cases that

00:41 do not support the proposition for which

00:43 they're cited, including three

00:46 quotations that don't exist.

00:49 My understanding is that you prepared

00:52 the order for new tri uh the denial

00:54 order for the trial court. Were those

00:57 citations in the version of the order

00:59 that you submitted to the trial court?

01:01 >> No, your honor. I do not believe so.

01:03 They were not. I did prepare an order.

01:07 That order was revised.

01:11 >> So those those non-existent cases were

01:13 cited in your initial brief opposing the

01:16 motion for new trial.

01:35 Your honor, I'm not aware

01:40 of that, but I would be glad to research

01:42 that.

01:46 >> That's a prosecutor in that clip.

01:47 someone who's responsible for sending

01:48 people to prison for a long time and her

01:50 office is relying on fake cases created

01:52 by an AI out of thin air. You have to

01:55 wonder how often this kind of thing

01:56 happens without being a detective. The

01:58 big law firm Sullivan and Cromwell, one

02:00 of the most famous and highric law enfor

02:02 law firms in the world, was just caught

02:04 doing the same thing, using AI to

02:05 deceive the court. Quoting from the New

02:07 York Times, uh, an elite Wall Street

02:09 Journal, uh, rather elite Wall Street

02:12 law firm has apologized to a federal

02:14 judge for submitting a court filing

02:16 replete with errors created by

02:18 artificial intelligence, including

02:19 hallucinations that fabricated case

02:21 citations. The AI generated errors came

02:24 in a recent motion in US Bankruptcy

02:26 Court in Manhattan. The firm provided a

02:28 ledger of the errors which spanned three

02:30 pages and totaled around three dozen.

02:33 Number of them involve the citation of

02:35 seemingly imagined passages from real

02:37 cases. So again, it's happening all over

02:39 the country. This is uh in Minnesota.

02:41 Watch.

02:43 In a recent court filing in Henipin

02:45 County Court, attorney Frederick Kak

02:46 wrote that prior Minnesota cases support

02:49 his argument, citing a 1992 case called

02:52 State by Sunquist versus Provos. The

02:55 problem, that case doesn't exist.

02:57 Neither did the other case connect cited

03:00 right before it or another case cited

03:02 later. When Judge Lorie Miller caught

03:04 it, she wrote, "The court wonders if

03:06 this citation may be the result of an AI

03:08 generated hallucination.

03:10 >> It's something that created to give you

03:11 a convincing answer of what you asked

03:13 for, but it's not necessarily an

03:15 accurate answer."

03:16 >> David Larson is a law professor at

03:17 Mitchell Hamlin, teaching a course next

03:19 semester on artificial intelligence and

03:21 the law.

03:22 >> Is this happening everywhere?

03:23 >> I think it's fair to say it's happening

03:25 everywhere. Okay. Um, one thing I

03:27 believe is that the temptation to use AI

03:30 is so strong that people just can't

03:32 resist it.

03:33 >> According to a database tracking AI

03:35 hallucinations in court filings, there

03:37 are 134 cases across the country in

03:40 which an attorney cited fabricated case

03:42 law. Carol 11 News found one other case

03:44 in Minnesota in July when Judge

03:46 Christian Sandy caught attorney David

03:48 Lutz citing a phony case. Lutz

03:51 confessed, admitting he used AI and

03:53 forgot to cross reference the cases.

03:58 Now,

03:58 >> all these lawyers should be arrested and

04:00 disparred immediately, but that's not

04:02 happening. Uh, the Georgia DA got a slap

04:04 on the wrist, and as you just heard,

04:05 these Minnesota attorneys simply have to

04:07 pay a fine and maybe take a scolding

04:09 from the state bar.

04:12 It's enough to make you wonder when

04:13 we're going to get full-fledged AI

04:15 attorneys who make a complete mockery of

04:17 the courtroom and get every case wrong.

04:19 Um, the more Sam Alman tells us that AI

04:22 super intelligence has arrived, the more

04:23 tempted these attorneys will be.

04:26 Actually, in one case in New York, this

04:27 has already happened. An elderly man

04:29 representing himself deployed an AI to

04:32 speak to the court on his behalf and uh

04:34 it it didn't really go well. Watch.

04:38 >> May it please the court. I come here

04:40 today a humble pro before a panel of

04:42 five distinguished justices.

04:44 >> Is this Hold on. Is that council for the

04:48 case?

04:49 >> That I generated that.

04:53 >> I'm sorry.

04:54 >> I generated that. That That is not a

04:57 real person.

05:01 >> Okay. It would have been nice to know

05:03 that when you made your application.

05:06 >> You did not tell me that, sir. I

05:08 received the application and you have

05:10 appeared before this court and been able

05:12 to testify verbally.

05:15 In the past, you have gone to my clerk's

05:18 office and held verbal conversations

05:20 with our staff for over 30 minutes.

05:23 Okay? I don't appreciate

05:27 being misled.

05:29 So either you are suffering from an

05:32 ailment that prevents you from being

05:34 able to articulate

05:36 or you don't you are not going to use

05:39 this courtroom as a launch for your

05:42 business sir. So if you are able to to

05:47 shut that off

05:53 uh well the AI lawyer's having a good

05:55 time. You can tell that it's safe to say

05:57 the AI did not in fact please the court.

05:59 Uh and and yet on and on it goes across

06:02 every industry. Uh this is from the

06:03 Atlantic in an article published just

06:05 the other day. Quote, "Earlier this

06:06 week, the New York Times reported that

06:08 the future of truth, Steven Rosenbomb's

06:11 much discussed book about how AI shapes

06:13 reality, contains more than half a dozen

06:15 fake or misattributed quotes. Rosenbomb

06:18 pinned some of them on his use of AI.

06:20 and he claimed responsibility for the

06:22 errors said he was investigating what

06:24 went wrong by the time I spoke with him

06:26 on Thursday though he was pointing his

06:27 finger elsewhere chat GPT effed up the

06:31 book Rosenbalm said it's been a rough

06:32 week for human authorship all around on

06:34 Monday a viral post showed a Nobel

06:37 novelist seemingly admitting to using AI

06:39 to sharpen her story ideas before later

06:41 claiming she had been misunderstood on

06:43 Tuesday allegations mounted that the

06:45 Trinidadian author Jamir Nazir had used

06:48 AI to write the serpent in the grove

06:50 which won the Commonwealth Short Story

06:52 Prize. By Wednesday, two of the other

06:54 five uh uh five prize winners had come

06:57 under similar scrutiny.

06:59 Now, seeing all these cases, you have to

07:01 ask yourself, when exactly are we

07:03 getting these superhuman intelligence

07:04 from these AI chatbots?

07:07 They can't write novels or legal briefs

07:09 without failing in spectacular fashion

07:11 and destroying careers in the process.

07:14 As I've always said, the technology is

07:15 extremely impressive and very useful in

07:18 many ways.

07:20 But there's a clear ceiling to the

07:22 knowledge base of these AI chatbots.

07:24 There's only so many books and websites

07:26 you can rip off before the AI has to

07:29 start thinking for itself and before

07:30 people stop voluntarily training these

07:32 models for free. If every leading AI is

07:35 already failing in significant ways,

07:38 then how much improvement can we really

07:39 expect? Exactly. This time of year is

07:41 always awkward because nobody's wardrobe

07:43 really matches the weather anymore. You

07:45 leave the house in the morning and it's

07:46 sometimes still chilly. By the

07:48 afternoon, it feels like summer and then

07:50 somehow by night need another layer

07:52 again. What most people really need are

07:54 clothes that are versatile enough to

07:55 handle all of the weather changes. And

07:57 that's part of the appeal of Mac Weldon.

08:00 Mac Weldon makes clothes that are

08:01 comfortable enough for everyday life,

08:03 but still look sharp and put together.

08:05 Things you can wear while traveling,

08:07 working, going out to dinner, running

08:09 errands, whatever the day turns into.

08:11 They offer updated basics like stretch

08:13 twill uh chinos, button-downs,

08:16 pullovers, and uh pieces from the Ace

08:18 collection. The Ace collection is

08:20 especially built for travel and long

08:22 days when comfort matters, but you still

08:24 want to look reasonably presentable.

08:26 Lightweight materials, flexible fit,

08:28 classic styles, clothes that move easily

08:30 between different parts of your day, and

08:32 the overall style is simple in a good

08:34 way. Nothing flashy or overly trendy,

08:36 just comfortable, well-made clothes that

08:39 work. Get moving with Mac Weldon. Go to

08:41 macweldon.com and get 20% off your first

08:43 order of $125 or more with promo code

08:46 walsh. That's m a ck weld d.com code

08:51 walsh. Now, as I mentioned when we

08:53 discussed AI data centers a couple weeks

08:55 ago,

08:56 uh I'm not suggesting that AI has no

08:58 valuable uses or that it can't improve.

09:01 AI can obviously benefit the country. It

09:03 can launch businesses and perform

09:05 valuable services. It can give certain

09:06 kinds of useful feedback. It can even I

09:09 mean it could it could diagnose medical

09:10 conditions sometimes more reliably than

09:13 many human doctors who are also using AI

09:15 anyway. Um it it can save lives

09:19 potentially given the right

09:20 circumstances.

09:21 But if we keep pretending that AI has

09:23 humanlike capabilities it never actually

09:26 has. If we keep threatening to destroy

09:29 millions of jobs and prevent young

09:30 people from even starting a life of

09:32 their own, then before long no one will

09:35 be able to use AI for any reason. There

09:38 will be a revolt that forecloses the

09:40 possibility and probably does a lot of

09:41 damage in the process.

Summarize any YouTube video instantly

Get AI-powered summaries, timestamps, and Q&A for free.

Generate your own summary →
More summaries →