Ok, I didn’t want to take the bait but this one’s just too much.
> “He also described a shift toward running AI locally rather than in the cloud – a move motivated by privacy, security, and the rising cost of inference as agents consume more tokens.”
Classic Apple. No more just beating the “security and privacy” drum, now its “tokens are expensive!”
<neanderthal voice/> Cloud scary. Cloud expensive. Mac good. Buy Mac!
> “He also singled out what he calls ‘transparent AI’ on iPhone and iPad, referring to features scattered throughout the operating system and third-party apps that work quietly without announcing themselves as AI.”
<neanderthal voice/> Apple use AI, Apple just not say it. Apple smart, not lagging behind industry! Buy iPhone!
How about you invest in developing your own models, correctly? And provide a secure and private inference cloud service on your fancy Apple silicon? And integrate that into your platform so Siri gets smarter without you farming queries out to Google Gemini? Bill me for it in iCloud+ I’ll probably pay for those tokens.
I see this kind of “let’s make candidates work for free” proposal from time to time.
It doesn’t work for software engineers because nobody wants to do free work. Also, free work isn’t consistently, and fairly testable because project requirements can evolve or completely change over a hiring period. Not to mention IP issues.
What has worked in my experience is a synthetic take home exercise that isn’t easily LLM solvable, the same one given to all candidates with the same constraints, and offering to be available to answer their questions by email. Sometimes we can tell a lot about the candidate just from the questions they ask, or how they package the solution, without even looking at the code.
In my experience (FANG hiring) less than 10% of candidates refuse to do this exercise, so it’s worked well for us.
I was already a Tesla owner and I reserved a Cybertruck right after I saw the original Cybertruck Unveil live stream on November 21, 2019. The infamous one where the window glass shattered.
That was when it was supposed to cost around $35,000.
Four years later when my reservation was ready to order, on December 8, 2023, the CyberTruck cost more than $100k.
Because it cost almost 3x more than what was originally advertised, I cancelled the order. I know many other people who canceled for the same reason. Keeping in mind this was after several delays, so I and many others with reservations were already frustrated with the product before it became available to order.
I can't put my finger on it but there's a weird tension between the two Dave's in this video. Almost like Rosenthal is trying to impress or earn the praise of Scherer.
I read the article, and while it’s great the model can generate relevant output- so what? The article doesn’t discuss any action being taken using that output.
Read through the Wiz MSA [0] at section 6 which discusses “Customer Data” and among other things specifically asks Customer not to send HIPAA data (perhaps to sidestep the issue you just raised) and concludes with this:
—
Customer hereby grants to Wiz a non-exclusive, worldwide, royalty-free right to use Customer Data to provide the Services and perform its obligations under this Agreement.
—
Or if reading terse legal documents isn’t your thing, go ahead and just read through Wiz’s own blog post about how their scanner works, which confirms they have full, direct access to customer EBS volume snapshots in the default “full SaaS” deployment model. [1]
Your point that due diligence would have taken issue with this might not be grounded in Google’s reality.
Google already have one of the best security teams in the industry - Project Zero [0]. They don't need Wiz's "enterprise" expertise for security.
This deal is about DATA. Wiz, as a cybersecurity vendor, have full remote access to their customers cloud compute storage (EC2 EBS volumes, etc) in the name of "security scanning" - this is actually part of their unique selling point - "agent-less scanning" which is unlike traditional security tools that require an agent installed in the OS. Instead, Wiz is able to just clone your full data volume and scan it locally in their cloud accounts/VPC.
With this deal Google has bought a ton of confidential data from Wiz's customers without their explicit knowledge or approval, and they will use it to improve Google's AI models like Gemini and probably several other products.
A year ago Google struck a $60M/yr deal with Reddit to exclusively license their content [1] for the same reason, and that data is probably much smaller and less valuable than the data Wiz has access to from their customers, which include companies like Morgan Stanley, DocuSign, Slack, Plaid, and others. [2]
“No punting — we can't keep building nanny products. Our products are overrun with filters and punts of various kinds. We need capable products and [to] trust our users.”
The irony of clicking a link to an article about “outrage fatigue” only to find it’s not even an article but a playboy style interview that goes on for several pages. Outrageous.
From Character.ai’s blog about safety measures they’ve taken since this law suit [1]:
“Over the past six months (…) we hired a Head of Trust and Safety and a Head of Content Policy and brought on more engineering safety support team members.“
Better late than never, but not by much. It’s not like they didn’t think about investing in safety before launch, they just didn’t prioritize it, to the public’s (and their brands) detriment.
I suppose this is why more legislation is needed in the AI space, because private companies can’t be trusted to prioritize public safety. Not everyone’s Apple.