Consulting firm quietly admitted to GPT-4o use after fake citations were found in August.
See full article...
See full article...
Deloitte will refund Australian government for AI hallucination-filled report
- Thread starter JournalBot
- Start date
"If you can't be part of the solution, there's good money to be made by prolonging the problem."
-- Despair.com poster titled "CONSULTANTS"
Upvote
12
(12
/
0)
A PARTIAL refund? I’d demand the entire amount back. Or they would lose all government business.
Upvote
19
(19
/
0)
Most consultants are trash anyways meant to push and rubber stamp some exec viewpoint in exchange for cash
Upvote
13
(13
/
0)
Yeah, I think there should be an all-new law based on it. Fraud could work, but only for the people who were paying AI for their services. It wouldn't do anything for the people that the AI cited in its output.
Though I think fraud requires intentional deception. I feel like this is more negligent deception (layperson using the word, so don't take that as the actual legal definition of "negligence.") I think these people really do think the AI would produce accurate output, or they would never pay for it and try to pass it off as legit.
The makers of AI, on the other hand, have more than enough evidence to know that any output by their software is likely to have errors. It feels like there's some possibility of making fraud stick there if they know that and still get people to use it.
Upvote
3
(3
/
0)
Deloitte, and the other "big 4" consulting companies aren't listed on any stock exchange. As auditors for a vast number of publicly traded companies, being publicly traded themselves creates massive conflicts of interest.
Upvote
9
(9
/
0)
There is a book, begging to be read by well versed American Citizens, that not only answers your questions but truly illustrates how deep seated the "corruption" is in AmerikKka
fascinating read too
FOLLOW THE MONEY
oNe
Upvote
11
(13
/
-2)
Almost half a million for a report? Wow...even if it was crafted by top of the field experts and took a couple months, thats a ridiculous amount of money to pay for a blame deflector.
Upvote
11
(11
/
0)
As Ars noted last June, "Yuck: Wikipedia pauses AI summaries after editor revolt". The most-used knowledge engine in the world found that AI summaries were not only worse than human-generated ones, they discouraged their human volunteers. Now Deloitte has found that its AI summaries are terrible, and they have damaged its own business. Trust is easy to lose and hard to earn. Who will hire them now?
They need to do serious crisis recovery. Scott Galloway likes to tell the story of the worst brand damage ever, that was also the best handled - the Tylenol cyanide poisoning case. A lunatic in Chicago started injecting potassium cyanide into drugstore boxes of Tylenol pills in 1982. Seven people died, and others from copycat crimes. The make of Tylenol, then Johnson & Johnson, did not simply clean out that drugstore, or that brand of Tylenol. They stripped all the Tylenol from all the shelves in the country and offered refunds to everyone. They developed tamper-resistant packaging, and the product turned out to be fine. Deloitte should take note!
They need to do serious crisis recovery. Scott Galloway likes to tell the story of the worst brand damage ever, that was also the best handled - the Tylenol cyanide poisoning case. A lunatic in Chicago started injecting potassium cyanide into drugstore boxes of Tylenol pills in 1982. Seven people died, and others from copycat crimes. The make of Tylenol, then Johnson & Johnson, did not simply clean out that drugstore, or that brand of Tylenol. They stripped all the Tylenol from all the shelves in the country and offered refunds to everyone. They developed tamper-resistant packaging, and the product turned out to be fine. Deloitte should take note!
Upvote
20
(20
/
0)
The AI does enhance the report. I mean the report content is probably as garbage as the citations but at least it cost the recipient nothing now lol, that is the enhancement
Upvote
2
(2
/
0)
Any non-AI report would’ve been littered with just as many errors.
The purpose of the report is to validate a prior worldview, not to evaluate the proposal.
AI is a godsend in this regard.
Upvote
-9
(2
/
-11)
The consulting industry is best understood as a way for the executive class to divert large sums of company resources to their friends and younger members of their class in exchange for applying the veneer of rigor to executive decisions. They not only don’t care how much of other people’s money they spend, it’s actually seen as better to spend a lot because when the business decision turns out to be flawed they have essentially prepaid for unlimited BS on demand to prevent accountability for the executives. In addition to the inexperienced recent grads, these companies maintain a stable of respectable senior guys of the right background who’ll show up in very nice suits and swear up and down nobody could reasonably have expected that pivot to blockchain to be anything less than a goldmine and that it’d be a major strategic error to factor the actual negative returns into someone’s bonus calculations. The overhead is paying for that service, too.
Upvote
20
(20
/
0)
There is a scene in my head that I'm 95% sure is from the Simpsons where someone commits insurance fraud and the insurance agent sent to pay out the money is so pure and trusting that he doesn't even think to question the payout despite increasingly obvious signs.
Eventually the evidence gets so overwhelming he has suspicions something is amiss and sets a trap, catching the culprit (mr burns maybe?) red handed and reads from a dictionary that it's textbook "insurance... Frowd???"
For the life of me I cannot find the scene, but it's been on my mind a lot (too much) lately.
Upvote
7
(7
/
0)
Usually, academic papers include an abstract written by the author. So, why would you need AI to supply a summary?
Upvote
16
(16
/
0)
The idea seems to be they get the "expert advice" for less than they'd pay an actual expert. The problem, of course, is if you're not an expert in the field you can't always recognize when your supposed expert, bet it so-called AI or just some asshat defrauding you, is spewing garbage. The managers and executives all think they're much more intelligent than the vast majority of them actually are, though, so good luck getting them to grasp this basic fact.
Upvote
8
(8
/
0)
Not slander, libel. It was published.
Anyway, you'd think, given the all-to-recent problem the previous government had with Robodebt, where an automated system erroneously determined that thousands of welfare recipients had been overpaid, everyone would be a bit more careful about checking anything that's been run through a computer. Although, I don't think anyone blamed Robodebt on an LLM, just bad policy, bad programming, and poor people management.
Upvote
9
(9
/
0)
The problem with big consulting is they never want to integrate with the clients. They just request a bunch of information then write a strategy document telling the client what they want. Then they implement the plan, provide two weeks of handover and piss off. The clients are left with some piece of shit platform that isn’t fit for purpose and cost 3 times what on prem teams could produce
/small company consultant for the last 15 years
We always joked that we loved big consulting. They gave us endless work fixing their bullshit implementations
/small company consultant for the last 15 years
We always joked that we loved big consulting. They gave us endless work fixing their bullshit implementations
Upvote
14
(14
/
0)
And that's the easiest mistake to spot. Earlier this year a friend of mine stated that ChatGPT has gotten "really good" at doing literature reviews with references and citations. Tried it, and not just were some references completely non-existent, most of the real references did not support the statements they were attached to.
Upvote
15
(15
/
0)
Fraud, as in common law fraud in the US, could be at play already because it the knowledge requirement generally (which can differ state by state) is Knowledge or Reckless Disregard: "False representations made recklessly and without regard for their truth in order to induce action by another are the equivalent of misrepresentations knowingly and intentionally uttered." Engalla v. Permanente Med. Grp., Inc Which is likely why they paid a refund to attempt to make Australia whole and to be able to point to it in a fraud suit as evidence that they were not trying to be reckless.
From your other post, a common type of damage to allege in defamation suits is reputational harm, and she could likely argue that her name associated with a fraudulent paper is reputationally damaging here in the US. Her higher bar might be showing that it was at least negligence to list her as the author.
Upvote
10
(10
/
0)
Successive Australian governments have hollowed out the public service in the name of saving money. Apart from the actual cost increases, the loss of institutional knowledge is crippling government departments at the state and federal level.
The fact that one of the big consulting companies is just feeding prompts to get an LLM to generate their work should be a wake up call to any competent opposition parties. But sadly, we don’t have any of those, just moronic yahoos shrieking “No!” At everything done by the government and jockeying for leadership of a party without a single real policy.
The fact that one of the big consulting companies is just feeding prompts to get an LLM to generate their work should be a wake up call to any competent opposition parties. But sadly, we don’t have any of those, just moronic yahoos shrieking “No!” At everything done by the government and jockeying for leadership of a party without a single real policy.
Upvote
14
(14
/
0)
Misciting reports to get the conclusion you want is exactly the main job of consultancy
Upvote
8
(8
/
0)
The right small consultant can be a godsend. A good friend was the CFO with a medium-sized company who consolidated divisions and laid him off (he saw it coming--no shenanigans or ill-will involved). As an interim gig, a vendor he used recommended him to another of the vendor's clients, whose business had been growing and was looking for consulting help setting up something more robust than his bookkeeper had been able to handle. My friend finished the analysis and made his recommendations, and the company's owner asked him to stick around and help implement it. "I spelled out the structure you need, and part of the recommendation is to hire a CFO instead of trying to do it yourself," my friend said.
"I'm offering you the job."
"I don't think you can afford me."
"Tell me your number."
He did, and they've continued to build that company, even through COVID.
Upvote
15
(15
/
0)
No no no, saying they "were made" alludes to a person doing something whereas mistakes actually just will themselves into existence.
Upvote
6
(6
/
0)
Exactly my question; how can the nonexistence of the supporting evidence for conclusion X have no effect on the veracity of conclusion X?
Confirmation bias as a service
Upvote
6
(6
/
0)
Has a consulting firm ever proven its worth in the past 20 years? I am asking genuinely. I remember an article and story basically summarizing they are just there to give businesses an "independent" perspective to tell the CEO what they wanted to hear (and what they were already wanting to do). Layoffs? Consult McKinsey and they will tell you "Our slide deck says layoff 10%. Do it!"
Upvote
3
(3
/
0)
Because consultancy companies are not there to provide consultations, they are an expensive exculpation tool. Nobody gets fired for hiring a consultancy company.
Upvote
7
(7
/
0)
But do you think you could show that just having her cited in a paper was damaging (especially when it was one of numerous citations) as opposed to something like claiming she was a co-author of the paper?
Upvote
1
(1
/
0)
"Don't worry -- we didn't write it" is certainly a slogan a company could have.The AI does enhance the report. I mean the report content is probably as garbage as the citations but at least it cost the recipient nothing now lol, that is the enhancement
Upvote
8
(8
/
0)
Because charging for said experts but not having to pay for them is even better greed always wants more for less.
Upvote
0
(0
/
0)
It seems you no longer get what you pay for.
This is no different to other scams where people sell you defective cars, ship products with half the content missing, etc.
This is no different to other scams where people sell you defective cars, ship products with half the content missing, etc.
Upvote
1
(1
/
0)
Having worked in an engineering management role in a company that was heavy on the use of consultants, I really wish this was the case.
If an executive has been with the company for some time, knows the full context of what is going on, and has a strong instinct for what the correct solution is, then that's probably not going to be a disastrously bad option. If they want to use a consultant to justify it, then that's a waste of money, but I can live with it.
However, that is rarely the case. Consultants are usually brought in when management has problems, doesn't understand them, has no technical ability to comprehend them, doesn't trust the solutions proposed by the company's own subject matter experts, and wants someone else to tell them what to do next.
Sure, the consultant will get the blame if it all goes wrong, but they're called in because they claim to understand any and all problems proposed, and to have a magic crystal ball that helps them understand the operational context of the business to ensure the solution is a right fit. Of course, this is nonsense... they know just as little as the management team does, but are paid to make confident recommendations despite that lack of knowledge.
Consultants are most heavily used by managers who are out of their depth, in companies where almost everyone is out of their depth, where most of upper management is clinging onto their jobs for dear life. In such companies, the executives know that's the situation, because they're in the same boat, so they put no credence in the actual judgement of their team and want to see external advice. It's a sign that the organisation is rich on head-count, low on talent, and any trust in staff to make correct decisions has evaporated.
Alternatively, consultants can be called in when a change is controversial, and management can't agree on a solution. In this case, the consultants aren't "told" to deliver a selected answer, but they're invariably leant on by different parties and given incomplete information by the various individuals that want the solution to land their preferred way. This is the worst case, because the consultant report, despite all its flaws, is used as a weapon.
Upvote
14
(14
/
0)
They consider 10% small?Kyle Orland said:
The problem isn't the size of the correction but the fact that we don't know how much AI was used in the production. Are the actual ideas sound? If they didn't bother to check something simple such as references, did they do any checking at all?
George Touche would never have allowed this to happen.
They mean producing more pages in less time (and with less staff). I can't wait until Detroit & Touch publish a genAI-enhanced cookery book with genAI recipes mixed in!
Upvote
4
(4
/
0)
Using “AI” to
Upvote
4
(4
/
0)
I sometimes think the only people hiring Deloitte are former employees who owe favors.. Look at the John Oliver episode - all they crank out is shit.
Guess who are former Deloitte consultants? Satya Nadella and Sundar Pichai.. Explains SO MUCH.
Guess who are former Deloitte consultants? Satya Nadella and Sundar Pichai.. Explains SO MUCH.
Upvote
8
(8
/
0)
Retrieval Augmented Generation (RAG) systems used to get data from documents into an LLM’s context are lossy.
OpenAI is coy about exactly how its document preprocessing works, but essentially it’s historically been necessary to try to extract relevant information prior to import because the context window wasn’t long enough. These data extraction methods have been criticised for lack of robustness. In particular if you look at the graph on the second page of that link, performance for getting exact matches for data is quite impressively bad.
So the entire paradigm of data retrieval from documents is a microcosm of the generative AI “slap together something that looks as if it works and then use marketing to paper over the cracks” business model and the outputs of these systems should never be trusted by serious people.
Upvote
7
(7
/
0)
Except for the very likely case that McKinsey and all other consultancies are doing the exact same thing as Deloitte did. After all, consultancies are simply made up of large number of MBAs that provide advice to industries in which they have absolutely no experience in getting paid by companies' C-suite to absolve themselves of the repercussions of the decisions they take by saying that they were guided by a world-respected consultancy.
Upvote
6
(6
/
0)
"We maintain confidence and trust in the accuracy of the report produced by the very fine people at this world-class consulting firm, now that a small number of corrections have been made and a few minor errors have been fixed. And in order to demonstrate our complete and total confidence in this outcome and in our choice of consultant, we have very clearly and obviously informed the public about the required corrections in a little box filled with a tiny font at the bottom of the Obituaries page. I can't remember what page number that is."
Isn't it grand when one incompetent organisation is deliberately shielded by another incompetent organisation due to the fear of backlash if too many people find out about the recent publicly-funded shitshow overseen by both.
edit:
From the Australian Financial Review (https://www.afr.com/companies/profe...or-slams-deloitte-s-ai-bungle-20251006-p5n0ch)...
James may not have explicitly instructed her staff to award the contract to Deloitte, but anyone who is familiar with office politics (or politics in general) and the "art" of managing people knows that there are plenty of ways to strongly influence such a decision without needing to be explicit about it.
This stinks, all of it.
Isn't it grand when one incompetent organisation is deliberately shielded by another incompetent organisation due to the fear of backlash if too many people find out about the recent publicly-funded shitshow overseen by both.
edit:
From the Australian Financial Review (https://www.afr.com/companies/profe...or-slams-deloitte-s-ai-bungle-20251006-p5n0ch)...
Bullshit.
James may not have explicitly instructed her staff to award the contract to Deloitte, but anyone who is familiar with office politics (or politics in general) and the "art" of managing people knows that there are plenty of ways to strongly influence such a decision without needing to be explicit about it.
This stinks, all of it.
Last edited:
Upvote
11
(11
/
0)
Upvote
1
(1
/
0)