Apple Bans Employees From Using ChatGPT Amid Its Own AI Efforts

[klasma]

I mean VERY generic, chore tasks like "write me a MacOS app showing a file selector dialog". It saves a lot of dev time but are still very generic. And MS can only know "this company is working on a MacOS app that shows a file selector dialog".

It’s a matter of scale. With 10,000 data points like this, you suddenly may be able to get a better picture of what the company is working on, in particularly if you can statistically cross-reference it with other company’s request data sets where you have additional information of what they’re doing.

And to take your example, in certain cases even just the information that you’re working on a macOS app can reveal information about your company strategy.

 

[neuropsychguy]

You're missing the big picture, but that's okay. I'll let you believe ChatGPT can't figure out what the asker is doing.

Some people are just too smart for people like us to understand.

 

[Tarzan of the Ocean]

whats next? CAD software? calculators? Not sure if this is the right approach apple. How do you plan to stay competitive in this rapidly changing industry DRIVEN by AI?

 

[vegetassj4]

ChatGPT ask Siri to discus with Jasper, the best way for YouChat to make sure that ChatSonic and Google Bard don't ask Socratic to prevent a new and improved Apple AI.

 

[laptech]

Wait till these AI bots start being used in discussions about criminal court cases, election results and political decision making with the output not being to people's liking, then the crap will hit the fan about how good or bad AI is.

 

[darngooddesign]

Makes perfect sense to me since we’ve already seen ChatGPT hacked with conversations exposed. This has less to do with Apple’s possible AI than it does with exposing confidential information to an AI which isn’t 100% private.

 

[bigboy29]

I work at a large software company everyone has heard of. I can promise you that each large company that has to report on compliance issues and works with customer data has the same policy in place and has communicated similar things to their staff.

It requires computing resources to stand up local LLM but it is doable for companies like this, and I am sure that they are doing that.

 

[darngooddesign]

Oh boy cant wait for Apple to be 5 years late and bring the worst ai to the party.

Tim cooks legacy is $ and nothing more

Tom Cook bad, Steve Jobs would never, Tim Apple something something. Does that cover all your talking points?

Why would Apple allow its employees to upload confidential information, which is what conversations related to its projects are, to a public AI which has already been hacked to expose private conversations.

Specific research related to its own AI plans is a different matter.

 

[PsykX]

If they don't want their secrets to be leaked through a prompt, OK this is understandable, but they should at least develop an internal ChatGPT using GPT-4, which is open source.

ChatGPT is so freakin userful and makes me so much more productive in my entire life, even professionally, it's ridiculous. If Apple employees can't use these tools anymore, they will be less productive than Microsoft and Google and will lose ground against them.

 

[toobravetosave]

Tom Cook bad, Steve Jobs would never, Tim Apple something something. Does that cover all your talking points?

Why would Apple allow its employees to upload confidential information, which is what conversations related to its projects are, to a public AI which has already been hacked to expose private conversations.

Specific research related to its own AI plans is a different matter.

Guess everyone ignores the part of the article where apple is building their own models huh

 

[hovscorpion12]

Oh wait, there's an official app for chatGPT now? I haven't noticed... I'll try that out when I get home :3

Anyways, huh. Not unexpected. It would probably lead to some of the secret juicy internal rumors being milled around in the models.

Yup. just got it downloaded and asked for 200-page Dissertation. AND IT DID IT.

 

[darngooddesign]

Guess everyone ignores the part of the article where apple is building their own models huh

Well of course they are.

However, that’s irrelevant to their not allowing the upload of confidential data to a non-secure, publicly accessible, and already hacked AI.

But Tim Apple bad right?

 

[toobravetosave]

Well of course they are.

However, that’s irrelevant to their not allowing the upload of confidential data to a non-secure, publicly accessible, and already hacked AI.

But Tim Apple bad right?

Yeah he p bad unless you’re a soulless shareholder

 

[darngooddesign]

Yeah he p bad unless you’re a soulless shareholder

Or he’s bad if you are a competitor who would like access to their stolen, confidential data if the public ChatGPT gets hacked again. Or you want to able to triangulate their plans once enough confidential datapoints have been uploaded to the AI database.

So why do you think they should upload any confidential data to a service which is outside of their control, a service which has already been hacked to expose private conversations and data? This is different from using it for tasks directly related to making a hopefully competitive AI.

 

[Ctrlos]

Totally appropriate for a business to control business-owned hardware and software used on their hardware.

Absolutely, but Apple take the secrecy a little bit too far. Imagine if Nintendo hadnt released *any* information, including admitting its existence of the new Zelda until a week prior to launch.

Sure it would sell but would you still get that launch day anticipation or the free marketing hype?

Apple used to keep their secrets because they were the leading light in the industries they sold products in. Since they aren't anymore I don't think the levels of secrecy they maintain are even needed.

 

[msackey]

This is a legitimate concern. ChatGPT is a souped up program that notes statistical correlation. It takes in a chunk load of written word as data to analyze statistical correlation (probability) between words and phrases. That's how it makes up sentences. It's not "smart".

It would be a danger then to feed company secretes to ChatGPT because that becomes raw data in which at some point it could spit out both true and false information using such raw data.

 

[ifxf]

Maybe chatgpt can create code with less bugs.

 

[crsh1976]

I’m not sure how Apple can enforce this rule. It’s a super useful program and unfortunately Pandora’s box has been opened.

It's common in enterprises to block sites, apps and web services/apis, ie social media sites are blocked where I work, as well as anything related to streaming, gaming or pornography, and chatgpt/openai derivatives as well since this went mainstream.

 

[darngooddesign]

I predict the next vectors in espionage will be uploading information to non-secure AIs through queries and using AI to retrieve and combine it at a later point. Makes complete sense for Apple to get ahead of that.

 

[VulchR]

ChatGBT is trained to be convincing, even when it lies. That's why most employees should avoid it - it crassly fabricates pseudo-facts that seem plausible. It might work with programming languages that have a very constrained structure, and few examples of intentionally erroneous information on the internet, but for virtually everything else ChatGBT does not distinguish among information, disinformation, and outright fabrication.

 

[arkitect]

Oh boy cant wait for Apple to be 5 years late and bring the worst ai to the party.

Tim cooks legacy is $ and nothing more

This is true…

To quote Liberace,"I cried all the way to the bank"

 

[montuori]

Given the well known issue of developers exposing private information in public Github repos, it's hardly surprising that firms are not allowing ChatGPT where the barrier to leaking information is even lower. Asking AI to provide a few example AWS API keys might be fun.

Furthermore, even if Apple (or some other company) could manage the leakage of their own code, there's the very real issue of licensed code. Firms like Apple no doubt have agreements with other companies about how 3rd party proprietary code must be handled and the consequences of breaking those agreements is usually pretty harsh. Asking an AI assistant to write a DB query against a proprietary schema (where relations and attribute names might be leaked) could well be grounds for a lawsuit. If you think this is an outlandish edge case, you'd be mistaken.

 

[Dydegu]

If Apple comes out with some sort of AI app, it’ll be out by 2057. And if it is in any way built on the Siri framework, that’s a hard pass.

 

[robbietop]

Totally agree. Ban it period. AI is dangerous, it encourages mediocrity

 

[Mackie924]

Makes perfect sense to me since we’ve already seen ChatGPT hacked with conversations exposed. This has less to do with Apple’s possible AI than it does with exposing confidential information to an AI which isn’t 100% private.

Yup. This is pretty much it. The ChatGPT security leaks more than justifies any company banning its use, IMO.