Expected future earnings. That first word often does a lot of heavy lifting in a nascent industry like AI. And yes, it probably does mean they'll need to raise their prices, but also that they'll be able to do just that and still stay in business.
Wow. Folks, we have a trifecta here. Someone who:
1. doesn't understand how it's possible to have high accuracy in a statistical average, even with noisy data;
2. doesn't understand how temporal fluctuations can average out over large time-scales to reveal longer-term trends; and
3. doesn't understand how to avoid effects due to sampling bias.
Maybe more than a trifecta. Anyone who wants to add to this list is welcome to do so.
How can a company be valued at $852,000 million dollars when they havenâ(TM)t made a cent of profit is beyond me.
Whenever you ask yourself that, just repeat in your mind: company valuations are based on expected future earnings.
My point stands. And I'm not confusing anything. The material covered in a course, and the level at which it is covered, are set before students enroll in the course. The pass/fail rate is determined by the kinds of students who enroll. It should not be set by departmental grading guidelines. How is a department to know whether a specific batch of students in a class in any quarter or semester will find the material difficult or not? They can base expectations for a specific course on historical statistics, but it's misguided to stipulate the percentage of fails by fiat as a supposed way of controlling how "difficult" a course is.
A passing grade in a course should indicate that you obtained competency with the material presented, at the level defined in the course description. You can't stop students from taking a difficult course they're not prepared to handle. Nor can you stop students from taking easier courses that they will ace because they're over-prepared. Pre-med students do this all the time, for example.
As for the topic in this story: failure rates are up because students are abusing AI tools and are ill-prepared for the course material. Sounds to me like the problem is a lack of adequate preparation for the incoming students. Prerequisites and/or skill-testing should be required of students before they can enroll in these courses. Perhaps the department would rather adjust the content of a course in order to keep a steady pass/fail rate. I think that's the wrong approach, because it just moves the goalposts, rather than giving students credentials for scoring goals on a standard playing field.
It's perfectly fine to have standards. And those who don't meet them should fail.
But on the other hand, those who do meet them should pass. So, I was surprised to read this in TFS:
The [Berkeley] electrical engineering and computer sciences department's grading guidelines state that 7% of students in lower division courses, including CS 10 and CS 61A, should receive D's and F's...
Not if they meet the standards of the course. Look, I know it's unlikely that you'll get an exceptional lower-division class where everyone is deserving of a pass. But it can happen, and "grading guidelines" should not force failing grades on people who met the course requirements.
In other news, Google is selling $80 billion in new stock to fund AI infrastructure. So, as I said in a post above, I suspect Google is only a temporary customer of SpaceX/xAI.
The fact that Google/Alphabet would risk stock dilution rather than just sell some bonds tells me that it sees a serious upside to the AI game, and needs flexible no-strings cash (i.e., without repayment schedules) to drive it forward. And the fact that it's renting GPU cycles from SpaceX means that it needs this compute-structure right now.
(Pepsi has canned Coke and vice-versa, for example.)
Hold on, I might be wrong about that. I remember hearing a news story about that several decades ago, but can't confirm it now.
[...] the people that actually know about business at SpaceX said "This Grok product is a big money pit, but there's a shortage of hardware that Google and Anthropic will pay loads to use"
Only temporarily, I suspect. Google and Anthropic are perfectly able to build their own compute-centers, and I suspect would rather do that than feed their competition. Once the hardware is available, of course.
Companies do help their competitors out from time to time when production-loads become unbalanced. (Pepsi has canned Coke and vice-versa, for example.) But, as I said, only temporarily.
You deserve a +1 Funny
And maybe a +1 Informative also.
"Spin doctors" literally exist for spacecraft, in the guise of Attitude and Control Systems (ACS) or Guidance, Navigation and Control (GNC) engineers.
You want some spacecraft to spin, in order to keep them stable and have their instruments sweep out 360-degree(*) coverage of the environment on every rotation.
(*) Or 4-pi steradian coverage, if you "spin" that way.
MAVEN is mentioned in the article you linked as a candiiidate for cancellation. Looks like there's no need now.
Many administrations have sought to cancel unmanned scientific missions, and boost whizz-bangy manned ones. But the current administration seems particularly focused on those objectives. I suppose it also helps their cause to cancel scientific missions with environmental objectives -- many of which are mentioned in your linked article.
Do you think they want to put humans back on the moon before China gets there? And before Trump's second term ends? Naaah...
The pussy Democrats will insist they live life in prison, on our dime. The rope would cost virtually nothing, but nope, let's set them up in house arrest on the taxpayer dime for the rest of their lives.
If cost is your argument, you picked a bad one. The full cost of trying, convicting, and sentencing someone to life in prison is cheaper than the full cost of capital punishment. Significantly cheaper, by a factor of about 2 to 5.
I'm pretty sure you'll argue that it costs so much money to execute someone because of all those pesky pre-trial procedures, elevated trial standards, mandatory appeals, and high-security incarceration of the condemned. Why not just get rid of all that? Well, because the state putting someone to death is a huge deal. It's a sentence for which there is no just remedy if it is arrived at incorrectly. So the state should be held to an exceedingly high standard for such trials, indisputably much higher than for life-in-prison cases.
IMHO, the state shouldn't bother. Don't hang 'em high. Lock 'em up instead.
When all is said and done I seriously hope some of [the people in this administration] will be found out to have colluded with foreign powers and hung for treason.
A few things:
1. We don't hang anyone anymore. Capital punishment typically involves lethal injection, although some jurisdictions may allow the condemned their choice. (Including hanging?)
2. Nobody can be convicted of treason if the USA is not in a war declared by Congress. Treason is a serious crime -- the only one mentioned in the US Constitution -- and the bar to be met for conviction is proportionately serious.
3. Under US law, collusion is not itself a specific, prosecutable crime. (But conspiracy is.)
4. I would like to see justice served as much as you, but not by killing people.
War is peace.
Freedom is slavery.
Ignorance is strength.
Bonus points to those who spot the reference.
Young people aren't getting hired. That's a problem. We need to find out what the cause is.
TFS/TFA have a new take on what that cause might be. I see that as possible enlightenment, not propaganda. A way forward to achieving a solution, though I'm not sure what it might be.
I mean most of these networks just tell you what you want to hear for the most part.
That's the problem. They can be obsequious and sycophantic, even when interlocuting with someone contemplating violence or suicide.
I have a hard time blaming any AI model at this point to be honest with you.
So do I. But it makes sense to blame the company that created the technology.
If the aborigine drafted an IQ test, all of Western civilization would presumably flunk it. -- Stanley Garn
