seraphsf

Future generation will be richer and better-off than the present. Saving your charity for the future is, effectively, stealing from the poor and giving to the rich.

Also, giving now maximizes the compounding effect of your charity. Saving 100 lives today is way better than saving 10 lives every decade for the next 10 decades.

What the test actually showed is that, given two conflicting goals from two human instructors, the model attempted to resolve the conflict by following one set of instructions, and subverting the other instructor.

It’s a good demonstration about how these models behave and what could go wrong. It is not an example of volition or sentience.

Frankly, the mission of "safe AGI" is better served by guaranteed wide distribution of the technology -- not only to MSFT but on fair commercial terms to all comers. I'm less worried about AGI being a bad thing, and more worried about AGI being monopolized by a powerful few.

Stunning.

2) in general, early Google used to hoard talent all the time. The founders would keep great people (or their friends) on payroll for ~ever just to have them stick around. That was most prevalent in the first decade of Google’s life, to my knowledge, and mostly applied to very senior people.

By the time Area 120 was pitched and approved (circa 2014), those days were largely gone. Area 120 was primarily filled with junior people (L4-L6) and constantly had to sing for its supper - it was not at all a sinecure.

A significant number of the people in Area 120 projects were folks who were stifled and/or wasted in their previous Google jobs. One explicit purpose of Area 120 was to prevent the loss of these entrepreneurs to outside startups. Not incidentally, this was a form of cultural reinforcement - Area 120 burnished Google’s reputation as a good home for entrepreneurial mindsets.

It’s not that hard to evaluate when something is working (ie the hard part in evaluation is false negatives, not false positives).

In Area 120’s case there was no coasting - if anything there was a hair-trigger standard to shut down underperforming projects.

I joined Area 120 with huge skepticism. It was hamstrung and inefficient in its own ways. And I agree it didn’t reach its potential - largely because it was encased in Google 2020 instead of Google 2007.

But to my surprise almost all of the projects were impressive, well-conceived, promising bets. And the people in Area 120 were among the top 10% of Googlers I worked with in my decade at the company.

Google killed Area 120 because of bureaucracy and politics, full stop. Google is worse off because of it.

And later: “In Mexico, the majority of businesses are small, informal, and off the books.”

It’s a tragic catch-22 of developing nations that governments are inept and corrupt for lack of funds. While distrust of government and rampant tax avoidance cause systemic underfunding.

Answer: typically, you’ll walk away with $5m (25%) or less. VC funds usually have a 1x preference, which means the get their $10m back (plus interest), and THEN they split the remaining proceeds with you 50-50%.

So if you take VC money, you might have to double your valuation just to keep your take-home value the same.

VC makes sense if you can grow fast and very large. But assuming you have scenarios to grow slower or to a smaller size, those scenarios often turn into bad ones if you’ve taken VC funding.

Levine’s article makes a great point:

“It really is wild that the SEC’s official position is now that it is illegal to “use unofficial communications to do things like cut deals, win clients or make trades.” “Conduct their communications about business matters within only official channels”! Imagine if that was really the rule! You can’t have lunch with a client and talk about business, or have beers with your colleagues and gripe about work, because that does not create a searchable archive for the SEC to review.”

I played around with the idea with my first Jeep Wrangler, a burly car with a perfectly exposed engine block. Never cooked anything, but I was endlessly entertained by the idea that I might. I suppose the transition to EVs will steal this particular daydream from my kids…

Carriers ask phone makers for config changes all the time. It’s possible this is a new capability that was requested by certain carriers.

To be fair, it’s also possible that the OP and I are misinterpreting what’s going on. For instance, iOS syncs your preferences across devices. Perhaps there’s some bug that’s causing the wrong setting to propagate back to this person’s phone because their iPads, etc are still set to allow “auto join”.

It’s intended as a consumer-friendly way to increase capacity in dense areas (like a sports stadium or mall) where the carrier’s cell towers don’t have enough capacity.

Wifi offloading is not new. AT&T helped invent these standards back in ~2009 when their network was getting crushed by massive increases in traffic as iPhone usage took off.

WiFi offload networks are configured as “Managed Networks” which are lower priority than any user-selected networks. You can disable them by turning off “auto-join”. (Also these WiFi offload networks are secure; you can’t spoof them).

However it appears that the original poster’s carrier (presumably Xfinity Mobile or Spectrum Mobile) has done something new - they’ve disabled the user’s ability to turn off “auto-join” on iOS. Some overzealous team is trying to lower their cellular costs. That’s because both Comcast and Spectrum rent capacity on Verizon Wireless towers, but their MVNO cellular service is not profitable unless their customers are using the cable company’s own WiFi fairly often.

However this (disabling “auto-join”) is a dumb move. It’s obviously problematic for users whose neighbors are broadcasting the [Xfinity WiFi or Spectrum Mobile?] SSID.

To my knowledge, no major carrier does this. If you’re on AT&T, T-Mobile, or Verizon, the “managed offload networks” can be easily disabled. And the major carriers are using higher-quality commercial WiFi networks for offload, not random home cable modems.

So, YMMV.