I found your substack a while back but to be honest I have not had a chance to read yet.
I was at the Nvidia GTC last week when Llion Jones from SakanaAI announced that "next day" they will have a big announcement about model merging.
I think this (Sakana's announcement) is a bigger deal than people realize - I am unsure how many people understand how much potential this has in the reasoning agent swarm type of usage where a lot of this work needs to be done. But compliance will be a beast to tangle if you merge a foundational "vanilla" model with a domain aware model that has baked in domain rules.
In regulatory and compliance areas, just being "compliant" is a huge task and very hard to do - I work a lot with foundational regulatory datasources (e.g. ITAR, DSM5, FAA regulations and what it means for companies down the supply chain to obtain compliance and use AI to help them "stay" compliant... that's difficult.
Thanks for writing about it - it jigged my memory and I went to read about it now. :)
I also read your job description and as a third party observer your job description sounds like you are looking for a "Deloitte consultant" type, but incidentally from the conferences I have attended in past year on AI, the large companies are definitely on the backfoot in their applied understanding of this whole AI paradigm shift.
AI needs to be baked into our lives, and most people working in compliance, regulatory and governance completely lack the comprehensive understanding how much of a paradigm shift this is. The field has moved way too fast - to some extent look at Gemini's DEI effort that ended in a public "walkback" from Google. The work needed is very nuanced...
The industry now focuses a lot on foundational models that are domain aware and partly on the chatbot interaction, with the massive application layer in the middle completely open and orphan in terms of an Ethical AI approach. (Vectorstore) Datasources and reasoning agents are missed completely, although everyone agrees that reasoning AI that is the next focus in 2024/2025.
Large companies like Google need to provide leadership and bake this into their toolset offering. you must provide the tools, not the actual governance - allow companies or governments to upload their own set of rules, policies, etc.
Anyhow, so, while I'm sure HR has its reasons to look for published authors (that's a fair, I get it), the job description seems to look for an MBA type with hopes they "get" AI... but empirical experience in past year tells me that person will continue to play catchup in understanding the true impact AI has on systems, peoples lives and how we can bake in ethical AI across regions, cultures, etc...
Nice, high quality substack.
I found your substack a while back but to be honest I have not had a chance to read yet.
I was at the Nvidia GTC last week when Llion Jones from SakanaAI announced that "next day" they will have a big announcement about model merging.
I think this (Sakana's announcement) is a bigger deal than people realize - I am unsure how many people understand how much potential this has in the reasoning agent swarm type of usage where a lot of this work needs to be done. But compliance will be a beast to tangle if you merge a foundational "vanilla" model with a domain aware model that has baked in domain rules.
In regulatory and compliance areas, just being "compliant" is a huge task and very hard to do - I work a lot with foundational regulatory datasources (e.g. ITAR, DSM5, FAA regulations and what it means for companies down the supply chain to obtain compliance and use AI to help them "stay" compliant... that's difficult.
Thanks for writing about it - it jigged my memory and I went to read about it now. :)
I also read your job description and as a third party observer your job description sounds like you are looking for a "Deloitte consultant" type, but incidentally from the conferences I have attended in past year on AI, the large companies are definitely on the backfoot in their applied understanding of this whole AI paradigm shift.
AI needs to be baked into our lives, and most people working in compliance, regulatory and governance completely lack the comprehensive understanding how much of a paradigm shift this is. The field has moved way too fast - to some extent look at Gemini's DEI effort that ended in a public "walkback" from Google. The work needed is very nuanced...
The industry now focuses a lot on foundational models that are domain aware and partly on the chatbot interaction, with the massive application layer in the middle completely open and orphan in terms of an Ethical AI approach. (Vectorstore) Datasources and reasoning agents are missed completely, although everyone agrees that reasoning AI that is the next focus in 2024/2025.
Large companies like Google need to provide leadership and bake this into their toolset offering. you must provide the tools, not the actual governance - allow companies or governments to upload their own set of rules, policies, etc.
Anyhow, so, while I'm sure HR has its reasons to look for published authors (that's a fair, I get it), the job description seems to look for an MBA type with hopes they "get" AI... but empirical experience in past year tells me that person will continue to play catchup in understanding the true impact AI has on systems, peoples lives and how we can bake in ethical AI across regions, cultures, etc...
Cheers - I'm subscribed !
Great comment, thank you. I agree EMM (potentially) has a lot of upside, but depends on whether it scales all the way up to the frontier & beyond.
On the role, probably not an MBA type (I don’t have one!), though more likely to have a policy research background.