>McKinsey—a consulting firm that works with ninety per cent of the Fortune 100
Ugh … these guys. When I worked for a Fortune 200 company as an EE, particularly in the late 90s / early 2000s It was almost that as engineers we couldn’t use the WC without McKinsey’s approval. It got pretty silly there toward the end of my tenure.
In fact I attribute the cancellation of a significant programme that we were 80 complete with to guidance from McKinsey provided to our marketing / executive management.
So, in a sense, perhaps … PERHAPS a well trained A.I. might be an improvement. But of course, I have to agree with Jaron Lanier on this one: there really is no AI, it’s just us. Maybe … given that AI can actually make a Bayesian calculation of probability of success/failure for a given decision by the AI before guidance is provided rather than, “… it is our guidance … that the appropriate course of action moving forward is to discontinue the programme …”
Interesting you bring this up. I went to an awards banquet this morning, and no fewer than a half dozen people I talked to at breakfast brought up this leak. I have to say I hadn’t read this carefully at the time of this writing. A cursory look would suggest that FOSS as an aggregate will be a dominant player in AI research and pushing the boundaries of what can be done and implemented. And because of severe budget constraint, the hardware will always be modest in the FOSS world.
Of course, we have always known that FOSS can outcompete projects developed inside the Cathedral.** A lot of what I read seems to be more of a panic around, “we are losing commercial opportunity to the FOSS community … OMG !!! …” This actually goes way back to the Microsoft Halloween documents (1999). This leak is the same thing, it seems to me.
One stat that jumped out at me was “LLMs on a Phone: People are running foundation models on a Pixel 6 at 5 tokens / sec.” But the FOSS community has always been able to produce tremendous efficiencies in hardware/software performance. Back in the mid-2000s a guy managed to put together a version of Linux that fit into a 50Mb footprint. It ran quite, efficiently in a very memory-limited profile. Damn Small Linux (DSL) is now discontinued, but I mention it simply because this case and many like it demonstrate how agile the FOSS community has been historically, how they will be so in the future, and how clever they are at finding novel solutions to solve their own personal problems/challenges. The same will apply to AI research and implementation.
But now that AI is in the hands of the FOSS community, there are all kinds of extremely positive as well as nightmare scenarios that could play out in the extremely near future. Since the rate of change is itself changing …
e.g., an A.I controlled drone swarm running on a Pixel 6 or other modest, mobile hardware built of commodity parts in the hands of a disaffected college student set loose on the institution of their choice. That’s seems like a plot line from a B-grade action movie, but it’s not that outrageous really. c.f., “Scalable Personal AI: You can finetune [sic] a personalized AI on your laptop in an evening” … from the leak.
When we really are in the soup is when CRISPR type technologies become readily available which may be controlled by an AI running on Pixel 6 class hardware and allows the same disaffected student to engineer their own viri. Moore’s Law is over. Don’t need brute force any more- The name of the game these days is AI finesse.
The FOSS community is complex in the sense that what it produces is unpredictable/emergent by nature and we really don’t/can’t know what’s just 60 days in the future. The Cathedral needs to budget/fund for the next quarter/shareholder conference call.
“The innovations that powered open source’s recent successes directly solve problems we’re still struggling with. Paying more attention to their work could help us to avoid reinventing the wheel.” I suppose this is the wrong time to mention that the wheel is constantly being reinvented- This is the nature of working in the Cathedral.
Mckinsey – into everything! ( valued over 10 billion owned by partners ) Not a single sector goes unnoticed by them. Big on data …big on AI technology. They are nicknamed -The Firm
I have not read the newest book yet – but considering it.
“When McKinsey Comes to Town: The Hidden Influence of the World’s Most Powerful Consulting Firm “.
cannot understand how this firm works with non profit sectors and governments. ..feels like a conflict of interest. (Do not think they are registered as lobbyists – yet get government contracts?)
makes me want to weep…
Global problems can’t be tackled if pay comes from a profit organization.
(ie a bank Ombudsman office that is payed by the bank – cannot get resolution re unjust/ethical issues- makes no sense – cannot be objective).
What is needed is a totally new model with the same depth of knowledge/insights, skills/expertise as one of these top consulting companies.
>McKinsey—a consulting firm that works with ninety per cent of the Fortune 100
Ugh … these guys. When I worked for a Fortune 200 company as an EE, particularly in the late 90s / early 2000s It was almost that as engineers we couldn’t use the WC without McKinsey’s approval. It got pretty silly there toward the end of my tenure.
In fact I attribute the cancellation of a significant programme that we were 80 complete with to guidance from McKinsey provided to our marketing / executive management.
So, in a sense, perhaps … PERHAPS a well trained A.I. might be an improvement. But of course, I have to agree with Jaron Lanier on this one: there really is no AI, it’s just us. Maybe … given that AI can actually make a Bayesian calculation of probability of success/failure for a given decision by the AI before guidance is provided rather than, “… it is our guidance … that the appropriate course of action moving forward is to discontinue the programme …”
Steve, did you see this: https://www.semianalysis.com/p/google-we-have-no-moat-and-neither
Interesting you bring this up. I went to an awards banquet this morning, and no fewer than a half dozen people I talked to at breakfast brought up this leak. I have to say I hadn’t read this carefully at the time of this writing. A cursory look would suggest that FOSS as an aggregate will be a dominant player in AI research and pushing the boundaries of what can be done and implemented. And because of severe budget constraint, the hardware will always be modest in the FOSS world.
Of course, we have always known that FOSS can outcompete projects developed inside the Cathedral.** A lot of what I read seems to be more of a panic around, “we are losing commercial opportunity to the FOSS community … OMG !!! …” This actually goes way back to the Microsoft Halloween documents (1999). This leak is the same thing, it seems to me.
One stat that jumped out at me was “LLMs on a Phone: People are running foundation models on a Pixel 6 at 5 tokens / sec.” But the FOSS community has always been able to produce tremendous efficiencies in hardware/software performance. Back in the mid-2000s a guy managed to put together a version of Linux that fit into a 50Mb footprint. It ran quite, efficiently in a very memory-limited profile. Damn Small Linux (DSL) is now discontinued, but I mention it simply because this case and many like it demonstrate how agile the FOSS community has been historically, how they will be so in the future, and how clever they are at finding novel solutions to solve their own personal problems/challenges. The same will apply to AI research and implementation.
But now that AI is in the hands of the FOSS community, there are all kinds of extremely positive as well as nightmare scenarios that could play out in the extremely near future. Since the rate of change is itself changing …
e.g., an A.I controlled drone swarm running on a Pixel 6 or other modest, mobile hardware built of commodity parts in the hands of a disaffected college student set loose on the institution of their choice. That’s seems like a plot line from a B-grade action movie, but it’s not that outrageous really. c.f., “Scalable Personal AI: You can finetune [sic] a personalized AI on your laptop in an evening” … from the leak.
When we really are in the soup is when CRISPR type technologies become readily available which may be controlled by an AI running on Pixel 6 class hardware and allows the same disaffected student to engineer their own viri. Moore’s Law is over. Don’t need brute force any more- The name of the game these days is AI finesse.
The FOSS community is complex in the sense that what it produces is unpredictable/emergent by nature and we really don’t/can’t know what’s just 60 days in the future. The Cathedral needs to budget/fund for the next quarter/shareholder conference call.
“The innovations that powered open source’s recent successes directly solve problems we’re still struggling with. Paying more attention to their work could help us to avoid reinventing the wheel.” I suppose this is the wrong time to mention that the wheel is constantly being reinvented- This is the nature of working in the Cathedral.
** this usage probably dates me. ref:
https://en.wikipedia.org/wiki/The_Cathedral_and_the_Bazaar?useskin=vector
Mckinsey – into everything! ( valued over 10 billion owned by partners ) Not a single sector goes unnoticed by them. Big on data …big on AI technology. They are nicknamed -The Firm
I have not read the newest book yet – but considering it.
“When McKinsey Comes to Town: The Hidden Influence of the World’s Most Powerful Consulting Firm “.
cannot understand how this firm works with non profit sectors and governments. ..feels like a conflict of interest. (Do not think they are registered as lobbyists – yet get government contracts?)
makes me want to weep…
Global problems can’t be tackled if pay comes from a profit organization.
(ie a bank Ombudsman office that is payed by the bank – cannot get resolution re unjust/ethical issues- makes no sense – cannot be objective).
What is needed is a totally new model with the same depth of knowledge/insights, skills/expertise as one of these top consulting companies.