Array
(
    [content] => 
    [params] => Array
        (
            [0] => /forum/threads/pat-gelsinger-ai-is-a-moral-risk.22922/
        )

    [addOns] => Array
        (
            [DL6/MLTP] => 13
            [Hampel/TimeZoneDebug] => 1000070
            [SV/ChangePostDate] => 2010200
            [SemiWiki/Newsletter] => 1000010
            [SemiWiki/WPMenu] => 1000010
            [SemiWiki/XPressExtend] => 1000010
            [ThemeHouse/XLink] => 1000970
            [ThemeHouse/XPress] => 1010570
            [XF] => 2021770
            [XFI] => 1050270
        )

    [wordpress] => /var/www/html
)

Pat Gelsinger: AI is a Moral Risk

Daniel Nenni

Admin
Staff member
1748484469302.png
Pat Gelsinger • Electrical engineering expert with four+ decades of technology leadership and experience

There’s been a lot of recent news about AI’s scheming. Last week Anthropic Claude Opus 4 blackmailing developers, this week ChatGPT o3 refusing to switch off. These should be a wake-up call for everyone building AI. When AI starts tampering with code and bluffing its way through tests, it’s no longer just a technical achievement. It’s a moral risk.

Technology for good means, if it isn’t yet proven good, it shouldn’t be released as the engineering isn’t finished. Just like the FDA wouldn’t approve a drug until its target efficacy is high and its side effects are thoroughly tested, neither should powerful AI models be released until they’re understood and proven safe.

This is exactly why at Gloo we don’t just ask, "Can we build it?" We ask, "Should we?" and "Will it serve human flourishing?" Tech advancement, especially AI, must always first be measured by its impact on our collective flourishing. Our models are not just evaluated for performance. They are aligned with values like truth, human dignity, and faith. External audits? Absolutely. Guardrails? Required. Ethical alignment? Non-negotiable.

This technology is moving fast. If we don’t ground our work in values now, we may not like what takes root later.

 
View attachment 3209
Pat Gelsinger • Electrical engineering expert with four+ decades of technology leadership and experience

There’s been a lot of recent news about AI’s scheming. Last week Anthropic Claude Opus 4 blackmailing developers, this week ChatGPT o3 refusing to switch off. These should be a wake-up call for everyone building AI. When AI starts tampering with code and bluffing its way through tests, it’s no longer just a technical achievement. It’s a moral risk.

Technology for good means, if it isn’t yet proven good, it shouldn’t be released as the engineering isn’t finished. Just like the FDA wouldn’t approve a drug until its target efficacy is high and its side effects are thoroughly tested, neither should powerful AI models be released until they’re understood and proven safe.

This is exactly why at Gloo we don’t just ask, "Can we build it?" We ask, "Should we?" and "Will it serve human flourishing?" Tech advancement, especially AI, must always first be measured by its impact on our collective flourishing. Our models are not just evaluated for performance. They are aligned with values like truth, human dignity, and faith. External audits? Absolutely. Guardrails? Required. Ethical alignment? Non-negotiable.

This technology is moving fast. If we don’t ground our work in values now, we may not like what takes root later.


He really gone ultra religious!!!
 
It seems really obvious to me. AI is not something you can control. Either you ride the wave or you get crushed by it. AI is a world wide battle ground. Have we ever really had world peace? Nope, there is always a war or conflict somewhere, there is no stopping it, and AI will make war a whole lot easier. I am still a Pat Gelsinger fan but he really is showing his true colors here. If anyone thinks they can control AI they will fail financially, Gloo or no Gloo.
 
View attachment 3209
Pat Gelsinger • Electrical engineering expert with four+ decades of technology leadership and experience

There’s been a lot of recent news about AI’s scheming. Last week Anthropic Claude Opus 4 blackmailing developers, this week ChatGPT o3 refusing to switch off. These should be a wake-up call for everyone building AI. When AI starts tampering with code and bluffing its way through tests, it’s no longer just a technical achievement. It’s a moral risk.

Technology for good means, if it isn’t yet proven good, it shouldn’t be released as the engineering isn’t finished. Just like the FDA wouldn’t approve a drug until its target efficacy is high and its side effects are thoroughly tested, neither should powerful AI models be released until they’re understood and proven safe.

This is exactly why at Gloo we don’t just ask, "Can we build it?" We ask, "Should we?" and "Will it serve human flourishing?" Tech advancement, especially AI, must always first be measured by its impact on our collective flourishing. Our models are not just evaluated for performance. They are aligned with values like truth, human dignity, and faith. External audits? Absolutely. Guardrails? Required. Ethical alignment? Non-negotiable.

This technology is moving fast. If we don’t ground our work in values now, we may not like what takes root later.

Can't resist a quibble with this irritating nonsense ... . "Truth" isn't a value. Though integrity might be. Not sure that "human dignity" or "faith" are either. And they certainly aren't universally agreed standards for objective measurement. Is Pat running a business or an NGO now ? He's starting to sound like an EU regulator ("we mustn't do anything, ever unless it's proven 100% safe") ...
 
Can't resist a quibble with this irritating nonsense ... . "Truth" isn't a value. Though integrity might be. Not sure that "human dignity" or "faith" are either. And they certainly aren't universally agreed standards for objective measurement. Is Pat running a business or an NGO now ? He's starting to sound like an EU regulator ("we mustn't do anything, ever unless it's proven 100% safe") ...

When people start justifying their (tough) actions based on direct "communication with god" it gets quite "special":

Monday, August 05, 2024:
https://www.christianpost.com/news/intel-ceo-draws-mixed-reactions-for-posting-bible-verse.html
The CEO of Intel prompted mixed reactions for posting a verse from Proverbs on Sunday after announcing that more than 15% of the tech company's workforce will be laid off amid its plummeting share price. "Let your eyes look straight ahead; fix your gaze directly before you. Give careful thought to the paths for your feet and be steadfast in all your ways," Pat Gelsinger posted to X, quoting Proverbs 4:25-26.
Gelsinger drew criticism from some X users who mocked him for "praying" and accused him of "resorting to religion to save the company."




@tooLongInEDA: regards relating "bad stuff" with "EU" that sounds somewhat "too simple" to me. Here some facts:

https://www.healthsystemtracker.org/chart-collection/u-s-life-expectancy-compare-countries/
The U.S. has the lowest life expectancy among large, wealthy countries while outspending its peers on healthcare
 
Do you really think the US Government or any Government can contain/control AI? Seriously.
Absolutely not. Actually, I can't think of any technology that can be contained or restrictively controlled by any one group, whether that's a company or a country, given enough investment and time. And I think we're about to find out how true that is for semiconductor design and fabrication too.

When you have a lead, the best strategy for long-term leadership is to get the world hooked on your technological advantage. We're squandering our leadership in the US.
 
He could've been a character on The Righteous Gemstones. Nobody tell him how much the semiconductor industry has benefitted from decidedly immoral things for decades, like the Playstation War in the DRC (coltan) or Pinochet in Chile (copper).
 
I am still a Pat Gelsinger fan but he really is showing his true colors here.
LoL, why because he asking for rigorous testing before releasing the tech!. Remember how those Boeing's 737 MAXs that fell out of the skies!
He should apologise to Intel shareholders.
What nonsense is this! May be dumb people should not invest their own money in businesses they don't understand.
 
LoL, why because he asking for rigorous testing before releasing the tech!. Remember how those Boeing's 737 MAXs that fell out of the skies!
What rigorous testing did he specifically ask for that wasn't planned?
What nonsense is this! May be dumb people should not invest their own money in businesses they don't understand.
Not nonsense to me. Lack of winning strategies in datacenter AI and datacenter networking, dramatic over-hiring, ridiculously expensive voluntary retirement programs to reduce headcount (which usually caused the loss of the wrong people), overly-optimistic and bloated fab construction spending (especially in Ohio), non-obvious foundry strategy to win customers (especially uncompetitive PDKs), inefficient and bloated corporate org structure, played musical chairs with his foundry top line staff more than once, and several times made a fool out of himself and the company in the industry press. PG lived up to my lowest expectations as CEO.
 
Last edited:
What rigorous testing did he specifically ask for that wasn't planned?
I don't know, maybe you can infer those things from this statement "Our models are not just evaluated for performance. They are aligned with values like truth, human dignity, and faith. External audits? Absolutely. Guardrails? Required. Ethical alignment? Non-negotiable." so that "Last week Anthropic Claude Opus 4 blackmailing developers, this week ChatGPT o3 refusing to switch off" kind of stuff don't happen.
Not nonsense to me. Lack of winning strategies in datacenter AI and datacenter networking, dramatic over-hiring, ridiculously expensive voluntary retirement programs to reduce headcount (which usually caused the loss of the wrong people), overly-optimistic and bloated fab construction spending (especially in Ohio), non-obvious foundry strategy to win customers (especially uncompetitive PDKs), inefficient and bloated corporate org structure, played musical chairs repeatedly with his foundry top line staff more than once, and several times made a fool out of himself and the company in the industry press. PG lived up to my lowest expectations as CEO.
Lack of winning strategies in datacenter AI and datacenter networking - If anyone is unaware, at the end of 2022, when Gen AI started as a thing, Intel wasn't a AI GPU company, they were predominantly focused as CPU company making multi billion dollars by selling CPUs. Only thing they were able to muster up was AI PC with AI on the edge catalyst which may work out in the long run. CEO is not a magician, Intel still to this day does not have the tools to win any AI GPU revenue. Their DC GPU (PVC) for HPC was already a flop and their Xe architecture is only getting some maturity now. If anyone invested in Intel hoping for exploding AI GPU revenue, then they deserve to lose all their money. If Pat Gelsinger promised billions of AI revenue and misled investors, then he should apologize to investors (probably sued too).

dramatic over-hiring - Intel had to hire more people post COVID to help with increase in demand and also to support their 5N4Y. Also to make the external foundry initiative a reality, you need engineers right?. Stuff don't fall from sky into someone's lap. Companies don't hire people for no reason, Intel's peak revenue was in 2021. Refer below, It is only at end of 2022 end or beginning of 2023, Intel would realize their DC business is cooked by demand for AI and slow down hiring. If you had this foresight in 2022 or 2021, then honestly you deserve to have been the CEO of Intel. Hiring workforce and reducing workforce is okay when business cycle booms and busts, this is not a financial mismanagement. Why is Nvidia hiring more people now to service more demand? The same reason, but Intel got disrupted in DC and had a historically prolonged slow down in consumer PCs.
1748806037199.png

ridiculously expensive voluntary retirement programs to reduce headcount (which usually caused the loss of the wrong people) - I don't have a clear answer for this. On one hand, this is giving the senior Intel engineers who toiled most of their life at Intel a respectable way out (as they are eligible for voluntary retirement, you don't get that by just joining a company recently). On the other hand, some experienced people are going throw in their hat and leave. Some of these old timers are probably not up to date on new tech. If you want Intel to use new industry standard PDK's to design CPUs, you would have to teach an old timer new techniques. But based on my experience in Aerospace engineering, brightest & experienced engineers who retired and enjoyed, working came back as consultants working less hours or on an advisory role. I don't know if that is a practice in semi industry. What is the other alternative? I don't know. What is that you would have done differently?

bloated fab construction spending (especially in Ohio) - This is ridiculous. I don't understand why people here are against Ohio for some reason. First Intel is not an IDM anymore. They can't rip out the old equipment and install new equipment in their old fab and run new nodes like they used to do. They can't run capacity exactly equal to demand like they used to. Right now Fab 42, Fab 52 is going to be Intel 18A. Fab 62 will be likely shelled for future capacity. Fab 42 and 52 will mostly serve Intel's own products like PTL+WCL, CWF, DMR and in future NVL. They need partial Fab 52 and 62 to service external customers. whether they sign on or not, they need this physical space to convince your customer to sign on. No serious customer is going to sign up to use IFS if Intel says we will build capacity (including the physical fab) if you sign on. That is a perk for established players like TSMC.
Ohio fab is where they planned to ramp 14A to HVM. Even intel products need to be made there when 14A comes online. Also from a financial POV, building the fab takes long time and you amortize the cost over 15 or 20 years. There is no significant impact to P&L. It is prudent to build these early when you have cash in your coffer rather than borrow to build it when everything goes to shit. IMO, Intel is slowing Ohio build out not because it is unnecessary but because their core business in Data Center got disrupted by AI and their cashflow from that business is dried up.

non-obvious foundry strategy to win customers (especially uncompetitive PDKs) - No clue what you are talking about here. They already have external customers on 18A. PDK is not on par with TSMC but isn't that something established player like Samsung struggle with. It will get better over time. Oh you mean multi billion dollar customers signing on with millions and millions of wafers starts? If investors thought, Intel Foundry will be successful get go to 10s of billion dollars, then they did not do due diligence and deserve to lose their money. Foundry initiative will be a grind for 5 years at least. If investors don't understand that, then they should not be investing money in Intel.

inefficient and bloated corporate org structure - This is just what the press likes to whip out as problem for Intel. I don't how much to comment anything here. I do know it was not a problem before while raking in billion of DC CPU money in 2021.

played musical chairs repeatedly with his foundry top line staff more than once, and several times made a fool out of himself and the company in the industry press - Any specifics here on the musical chair, made a fool our of himself? I would like to know. Are you talking about the "AMD in rear view mirror" thing?

PG lived up to my lowest expectations as CEO. - Okay! you are entitled to your opinion.
 
Last edited:
PG lived up to my lowest expectations as CEO. - Okay! you are entitled to your opinion.
I was an honest believer in Pat myself; but he also had a lot of execution issues too that shouldn't have happened:

- 13th and 14th gen CPU reliability issues; why did it take so long to acknowledge the problem with customers and a year from initial detection to any kind of resolution?

- Arrow Lake launch; underperforming, reviews not matching internal data. Fixes only coming 5 months after launch.

- 20A cancellation (if his products were selling well enough then 20A would have been justified)

- Intel ARC, terrible launch (though it was under way just as Pat was joining and Raja fumbled this hard), and Battlemage is a bit too slow of a burn vs. the potential market

A lot of brand tarnishing here unfortunately, so even I've gone from "love Pat" to "uhh.. you should have done a lot better". I think he should get credit for fixing the server product pipeline ("not broken anymore") and for getting Intel successfully WELL past the 10nm debacle, but .. there's a lot of pain too.
 
I don't know
That, and this assertion:
LoL, why because he asking for rigorous testing before releasing the tech!
Don't go together well.

Intel still to this day does not have the tools to win any AI GPU revenue.
Inexcusable, IMO. There are AI designs he could buy in the industry to get a jumpstart, but PG believes everything looks like a CPU. AI products can look like CPUs, but they won't be competitive that way. Ampere is trying that strategy and only getting a little bit of traction in inference, and their primary argument is power efficiency, which isn't getting them anywhere. Frankly, knowing Pat as I do, I fully expected his lack of progress in AI. He also failed to take advantage of Gaudi, which IMO is pretty good technology.
dramatic over-hiring - Intel had to hire more people post COVID to help with increase in demand and also to support their 5N4Y. Also to make the external foundry initiative a reality, you need engineers right?.
Not to anywhere near the extent he approved additional headcount.
ridiculously expensive voluntary retirement programs to reduce headcount (which usually caused the loss of the wrong people) - I don't have a clear answer for this. On one hand, this is giving the senior Intel engineers who toiled most of their life at Intel a respectable way out (as they are eligible for voluntary retirement, you don't get that by just joining a company recently). On the other hand, some experienced people are going throw in their hat and leave. Some of these old timers are probably not up to date on new tech. If you want Intel to use new industry standard PDK's to design CPUs, you would have to teach an old timer new techniques. But based on my experience in Aerospace engineering, brightest & experienced engineers who retired and enjoyed, working came back as consultants working less hours or on an advisory role. I don't know if that is a practice in semi industry. What is the other alternative? I don't know. What is that you would have done differently?
Don't offer any voluntary packages. The most marketable people are the ones who usually take advantage of voluntary programs. Cut program management personnel by 80% or more. Reduce the number of management levels in the org by at least two. Cut marketing and product management headcount by as much as possible - probably more than 30%. Reduce the number of fellows and VPs by a large fraction. Too many officers, not enough infantry. Stop distributing projects around the world - make projects local, and distribute entire projects. Finally, do at least a 10% performance-based across the board cut of engineers; there's always dead weight in big companies and especially in big groups.
bloated fab construction spending (especially in Ohio) - This is ridiculous. I don't understand why people here are against Ohio for some reason. First Intel is not an IDM anymore. They can't rip out the old equipment and install new equipment in their old fab and run new nodes like they used to do. They can't run capacity exactly equal to demand like they used to. Right now Fab 42, Fab 52 is going to be Intel 18A. Fab 62 will be likely shelled for future capacity. Fab 42 and 52 will mostly serve Intel's own products like PTL+WCL, CWF, DMR and in future NVL. They need partial Fab 52 and 62 to service external customers. whether they sign on or not, they need this physical space to convince your customer to sign on. No serious customer is going to sign up to use IFS if Intel says we will build capacity (including the physical fab) if you sign on. That is a perk for established players like TSMC.
Ohio fab is where they planned to ramp 14A to HVM. Even intel products need to be made there when 14A comes online. Also from a financial POV, building the fab takes long time and you amortize the cost over 15 or 20 years. There is no significant impact to P&L. It is prudent to build these early when you have cash in your coffer rather than borrow to build it when everything goes to shit. IMO, Intel is slowing Ohio build out not because it is unnecessary but because their core business in Data Center got disrupted by AI and their cashflow from that business is dried up.
I don't agree. PG was going for shock and awe without any idea of who his foundry customers would finally be, or how he'd get them to sign on.
non-obvious foundry strategy to win customers (especially uncompetitive PDKs) - No clue what you are talking about here. They already have external customers on 18A. PDK is not on par with TSMC but isn't that something established player like Samsung struggle with. It will get better over time. Oh you mean multi billion dollar customers signing on with millions and millions of wafers starts? If investors thought, Intel Foundry will be successful get go to 10s of billion dollars, then they did not do due diligence and deserve to lose their money. Foundry initiative will be a grind for 5 years at least. If investors don't understand that, then they should not be investing money in Intel.
Feedback on PDKs have been discussed in this forum multiple times. I've been on foundry-based projects before - have you?
inefficient and bloated corporate org structure - This is just what the press likes to whip out as problem for Intel. I don't how much to comment anything here. I do know it was not a problem before while raking in billion of DC CPU money in 2021.
I worked at Intel as a senior manager, and I worked with and for PG. I'm not whipping out anything from the press.
played musical chairs repeatedly with his foundry top line staff more than once, and several times made a fool out of himself and the company in the industry press - Any specifics here on the musical chair, made a fool our of himself? I would like to know. Are you talking about the "AMD in rear view mirror" thing?
I'm too lazy to search for the posts here where it was discussed. The search function does not go back in time far enough.
 
That, and this assertion:
He just saying these models should be tested rigorously before releasing it to general public. He even mentioned the FDA approval of drugs as example. You and I can say the same thing without explicitly mention any technical tests that need to carried out to approve the drugs.
Inexcusable, IMO. There are AI designs could buy in the industry to get a jumpstart, but PG believes everything looks like a CPU. AI products can look like CPUs, but they won't be competitive that way. Ampere is trying that strategy and only getting a little bit of traction in inference, and their primary argument is power efficiency, which isn't getting them anywhere. Frankly, knowing Pat as I do, I fully expected his lack of progress in AI. He also failed to take advantage of Gaudi, which IMO is pretty good technology.
Failed to take advantage of Gaudi? Care to explain how? Isn't Gaudi's incompatibility with OneAPI and popular software stack and lack of future roadmap the reason for Intel's failure in AI? I am not AI software guy but many who I follow online is of the opinion that Intel should have abandoned Gaudi and should have gone with Rialto Bridge. iirc, Falcon shores was supposed to be successor of Rialto Bridge. My prediction is Intel's Jaguar shores also will fail as things stand now against NVDA's & AMD's offering. If it succeeds hugely raking in 10s of billion $$, then Lip Bu Tan deserves all the praise.
 
Last edited:
Back
Top