Why AI Still Lacks True Agency — And What Founders Must Build Instead

Summary

Liftoff with Keith Newman – Episode Summary

Episode Overview

Title: Why AI Still Lacks True Agency — And What Founders Must Build Instead
Date: January 27, 2026
Guest: Archever Blackwell, PhD – AI researcher, author, and former Ask Jeeves exec
Host: Keith Newman

Keith Newman dives into the current and future state of AI with Dr. Archever Blackwell, discussing why AI lacks genuine agency, the pitfalls of current AI integration in business, and where the best near-term opportunities lie for founders. Blackwell draws from his academic background, recent book ("The Creative Agency’s Guide to AI"), and hands-on experiences to provide founders with actionable advice for building in the current AI era.

---

1. Introduction & Background (00:38–04:39)

Main Points:

• Dr. Blackwell shares his lineage in AI, having studied under UCSD pioneers Jeff Elman and Liz Bates, whose work on neural networks laid groundwork for today’s AI revolution.
• The 1980s-90s academic debate: symbolic AI (Chomsky) vs. connectionist approaches (Elman, Bates). Data-driven, neural-network models have now triumphed as computing power caught up.
• Many theoretical foundations from the past still apply to modern AI models—showing both technological continuity and the importance of grounding in real data.

Notable Quote:

“They stayed close enough to the data that I think it really helped to produce a lot of what we’re seeing today, which is very, very close to what they amazingly came up with, you know, way back in the day.”
— Archever Blackwell (03:53)

---

2. AI’s Societal & Economic Impact: Is Hype or Doom Justified? (04:39–10:37)

Key Insights:

• Blackwell positions himself between “AI doomers” and techno-optimists. He emphasizes parallels with past technological disruptions (printing press, telegraph, computer, internet): each brought upheaval, but society ultimately adapted.
• Worries about job losses are real, but new jobs, roles, and adaptation always emerge—though with pain and transition.
• Modern generative AI (ChatGPT, Claude, etc.) have only been widely available for ~3 years. It's too early to judge the “productivity miracle” or real impact, just as it would be in 1996 for the web.
• Sophisticated outputs aside, current AI “thinking” isn’t true consciousness or self-awareness.

Memorable Moment:

“Every really amazing technological innovation ... all of those have produced really great progress, but also really great upheavals in society.”
— Archever Blackwell (05:05)

On AI Consciousness:

“Whether or not that means that they’re actually conscious and aware I think is a relatively open question. But I think at this point the answer is pretty much no, but not completely rule outable...”
— Archever Blackwell (08:29)

---

3. Building with AI: Pitfalls and Best Practices for Founders (10:37–14:31)

Discussion Points:

• Too many companies “bolt on” AI—creating clunky, disconnected processes that fail to actually boost productivity.
• Integration must be seamless and embedded into existing workflows.
• Lessons from the rise of enterprise IT: companies that best integrated and serviced new tech captured outsized value.
• As the ecosystem matures, more robust, easier-to-integrate AI infrastructure will emerge.

Notable Quote:

“One of the biggest problems ... is that the implementation doesn’t integrate into the existing workflow ... not the fault of the AI itself, that’s the fault of the way it’s been integrated in.”
— Archever Blackwell (11:09)

---

4. Opportunities & What’s Next: Local LLMs, Verticalization, and Co-working AI (14:31–23:22)

Major Trends Identified:

• Hand-in-glove Collaboration: The future is not just asking AI for outputs, but true iterative collaboration (“co-workers,” e.g. Anthropic’s Claude CoWork).
• Local Large Language Models (LLMs): Blackwell sees enormous potential for smaller, local models—especially in privacy-sensitive fields (agencies, medical, legal). Tuning these models with proprietary data preserves “voice” and confidentiality.
• Vertical & Local Models: Movement away from generic “Swiss army knife” AI toward sector-specific, enterprise-customized LLMs. These are cheaper, more private, and increasingly powerful.

Quotes:

“There’s an enormous amount of potential in using local large language models ... and they’re remarkably powerful for their size.”
— Archever Blackwell (16:21)

“Most people, when they think of AI right now, they think of the frontier models ... But there’s an enormous amount of potential in using local large language models...”
— Archever Blackwell (17:27)

“The problem a lot of people have with using AI for creative work is everything kind of comes out sounding the same. And you want to preserve the voice of your agency, the voice of your product.”
— Archever Blackwell (20:02)

---

5. What AI Can & Can’t Do Creatively: Judgment, Not Just Generation (21:46–23:22)

Key Points:

• AI’s creative output too often sounds generic; but localized, tuned models can maintain an agency’s distinct “voice.”
• The biggest value for creative agencies may lie in AI’s ability to judge and target content, not just write it.
• Practical example: Having AI analyze which creative assets will best perform for a given Facebook audience, reducing trial-and-error cycles.

Quote:

“AI can also ... give you some judgment on content. And so what you can have it do is say, hey, for this particular kind of content, this is who I think you’re most likely to be successful targeting.”
— Archever Blackwell (22:32)

---

6. No AGI, Just Really Smart Tools (23:22–27:58)

Blackwell’s View:

• Makes no strong predictions about AGI (artificial general intelligence—a system with human-level awareness and independent goal-seeking).
• AI will show increasingly sophisticated behaviors, but will remain fundamentally passive tools—more like Star Trek’s computer than a character with agency.
• True agency may come only when AIs are embodied and able to learn continuously; but for now, and likely for years, these systems execute tasks when asked but don’t want anything.

Quote:

“The computers never kind of take over ... they’re not ever like characters in the story or anything ... they’ll still essentially be tools waiting for us to interact with them.”
— Archever Blackwell (24:46)

---

7. Advice for Founders & Builders: A New Playbook (27:58–30:32)

Actionable Advice:

• The old “go take a course” model doesn’t cut it in an environment changing this fast; founders and builders need to take charge of their own education, experimenting daily with AI and learning to harness it as a tool.
• Skill focus is shifting: what matters is an ability to reason with AI, optimize prompts, and integrate output—not just to code.
• Founders must be “super motivated” and proactive to stay ahead of the next wave.

Memorable Quote:

“You have to really be super motivated and be a self-starter ... I’m going to go in there, I’m going to interact with these systems on a regular basis.”
— Archever Blackwell (29:23)

---

8. Blackwell’s Current Projects & Final Thoughts (30:54–34:17)

Key Points:

• Blackwell is developing a local LLM platform for creative agencies and potentially other privacy-sensitive industries.
• System allows agencies to train the model on their own material, keeping everything air-gapped and confidential.
• The interplay of privacy, voice distinction, and practical workflow are central to uptake.

Resources Mentioned:

• Blackwell’s newsletter: [InsideTheBlackBox AI]
• Application: [YourVoiceCraft AI]
• Book: “The Creative Agency Guide to AI”

Closing Sentiment:

“The local language model is going to be for a lot of people where their buy-in to AI starts with, or at least is an important component.”
— Archever Blackwell (33:29)

---

Notable Quotes & Timestamps (Quick Reference)

• Staying close to data:
  • “They stayed close enough to the data ... very, very close to what they amazingly came up with, you know, way back in the day.” (03:53)
• On societal adaptation to tech:
  • “Every really amazing technological innovation ... have produced really great progress, but also really great upheavals in society.” (05:05)
• On integration failings:
  • “Not the fault of the AI itself, that’s the fault of the way it’s been integrated in.” (11:09)
• Local LLMs as the future:
  • “There’s an enormous amount of potential in using local large language models ... for their size.” (16:21)
• Judgment over just content creation:
  • “AI can also ... give you some judgment on content....” (22:32)
• AGI skepticism:
  • “They’ll still essentially be tools waiting for us to interact with them.” (24:46)
• On founder mindsets:
  • “You have to really be super motivated and be a self-starter...” (29:23)

---

Timestamps for Key Segments

• 00:38–04:39: Blackwell's academic roots and early AI history
• 04:39–10:37: AI and societal disruption, productivity, and consciousness
• 10:37–14:31: AI integration mistakes and lessons for business
• 14:31–19:07: The promise of local vs. frontier LLMs; privacy, cost, and tuning
• 19:07–23:22: Verticalization, creative agency AI, judgment, and targeting
• 23:22–27:58: AGI discussion and philosophical boundaries
• 27:58–30:32: Advice for founders and the future of technical education
• 30:54–34:17: Blackwell’s projects, final thoughts, and resources

---

Summary Takeaway

Founders should focus on building tightly integrated, local, and data-tuned AI tools, especially for creative and privacy-sensitive sectors. Forget the sci-fi AGI hype—real innovation lies in smart, collaborative applications and a new personal playbook centered on continuous learning and hands-on experimentation.

---

Host sign-off:

“This is a story that just continues to unfold at a very fast pace.” (33:56)

Transcript

[Podcast Sponsor Announcer]

Foreign. This episode is brought to you by the outstanding team at Compass Strategic Advisors, your trusted partner in driving strategic growth. Whether it's expanding into new territories, launching products, rebranding, or hiring key staff and board members, they offer the expert guidance every step of the way. With a versatile approach and a proven track record, Compass is the go to resource for both startups and established companies looking to scale with confidence. Thanks and show your support for this podcast by sharing some sponsor appreciation. Check these guys out at bww.compass-strategic-advisors.com well.

[Podcast Host]

This is super exciting for me because I, I get a lot of active practitioners in the startup game and sometimes we get somebody here who's seen it all, been there and done that, and is is kind of going to keep us ahead of the curve, as it were. Archever Blackwell, PhD, long involved in the AI world. Welcome to Liftoff.

[Archever Blackwell, PhD]

Thank you. I'm very, very pleased to be here.

[Podcast Host]

It's a great pleasure to have you. You're down in Los Angeles today, right?

[Archever Blackwell, PhD]

That's right. Where it has right now, what, 65 degrees, I think.

[Podcast Host]

And let's make sure our friends on the east coast of the United States and the world understand it is just humming down in Southern California these days. So many exciting companies and business. And then you've got the great weather. Let's just keep all that, all that fear of, of earthquakes and fires out of the picture as long as possible.

[Archever Blackwell, PhD]

Exactly.

[Podcast Host]

Well, to dig right into it, you have a new book, the Creative Agency's Guide to AI. I'm excited to dig in because you've been there and done that for a while. You studied under Jeff Elman and Liz Bates, right?

[Archever Blackwell, PhD]

That's right.

[Podcast Sponsor Announcer]

Beginning.

[Podcast Host]

Right.

[Archever Blackwell, PhD]

Yeah. They were two of the big pioneers. Ucsd UC San Diego, which is where I got my doctorate, was really like the hotbed in the 80s and 90s for a lot of what was called then neural networks and connectionism and is now what we're calling AI and transformers and LLMs and vision models. And Liz Bates and Jeff Elman were. But when I go back and read some of their papers now, 30 or 40 years later, I'm really amazed at how much of what's going on now is applicable to what we're doing right now and how really they just got so much right. And at the time, the problem, of course, was computers were much more underpowered compared to what we've got today. I mean, what we have today is amazing. Also the algorithms and our understanding of how language worked and of how these neural network Style algorithms work that are, you know, they're kind of vaguely based on how the brain works, but not exactly. And so what we're seeing today, 40 years later, is really kind of the fruition of a lot of what their vision and that vision. It's interesting because at the time, that vision was very much a controversial one and it was not in the mainstream at all. The sort of mainstream in psychology and psycholinguistics was what Noam Chomsky, who's still with us, really kind of his view of what language was like, which was much more ironically, machine like and computer like, but not the way we think of AI, but much more symbolic AI. And they really had to fight. And I have to say I admire them greatly for not backing down and for really pushing their views and saying no. And what I think that was all coming from the fact that, that they were looking at the actual data. They were looking at what, how children were learning language. They were looking at how humans used language once they fully acquired it. They were looking at how people were multilingual and how people who had brain damage were using language or were not using language. And they stayed close enough to the data that I think it really helped to produce a lot of what we're seeing today, which is very, very close to what they amazingly came up with, you know, way back in the day.

[Podcast Host]

Yeah, A lot of conversation around the thinking machine and how far we are going to take that.

[Archever Blackwell, PhD]

Right.

[Podcast Host]

The idea of, of AI starting to interpret now, starting to think on its own. Where do you chart that when you look at these trends?

[Archever Blackwell, PhD]

Well, it's funny because I've just been working on a paper about how conscious is AI and I'm very much in the middle on the hype doom spectrum. So I don't think that a lot of the doom is warranted, but I do think that we're seeing significant economic disruption because of AI, and I think we're going to see a lot more of it. But I mean, I would say keep in mind that every really amazing technological innovation, from the printing press to the telegraph to the train to, you know, television and radio and then the computer and then the Internet, all of those have produced really great progress, but also really great upheavals in society. Right. And I think that if we take all of those past historical events as a guide, what you're going to find is that we did adapt to those and there was some pain. I'm not going to deny that because there definitely was some pain, but I think there was a lot of concern. For instance, you had all these people in back offices in the 40s and 50s. You had these huge rooms full of people at those old style know calculators. And the concern was, oh, all those people are going to be out of work. What are they going to do when we get these computers in that take the role of everybody in this room basically. But somehow there was an adaptation and there are still plenty of jobs out there that humans can only do that AI can't do. And I think it's going to be a long time before we see a problem where there's just like nothing left for human beings to do. I mean that may eventually happen like in some Star Trek future universe, but I don't think that's anytime soon.

[Podcast Host]

There are, there is a big delta now between those who believe we're seeing significant productivity gains today.

[Archever Blackwell, PhD]

Yeah.

[Podcast Host]

And those who don't see it as well. And we've seen reports supporting those points. And then we are also seeing the future cast of what jobs you get eliminated. Where do you want to go run to to maintain your, you know, economic independence, you know, your work satisfaction index and all of those elements. So how are you starting to, to track all of this?

[Archever Blackwell, PhD]

Well, so the first thing is we, I think that we are seeing some impacts of it. We're not necessarily seeing them, you know, in sort of the leading economic indicators yet. But I would point out, look, it's really early, I mean chatgpt and those kinds of systems have only been out for what like about three years in sort of their current incarnation. And I remember seeing the first web page probably that was what, 1993 when and that was an academic web page. I mean that wasn't even like, you know, Banana Republic or something. And three years later they, the Internet was still very much in its infancy and yet it was pretty clear I think even then that it was going to be a huge phenomenon that websites were eventually going to take over a lot of E commerce and that people were going to be using the Internet for so many things that they had previously been, you know, running to the store to do or calling up someone on the phone to do or something like that. So I don't think that we've really had enough time to, to see the full impact of what's going on with these systems. And I'm, when I see, I use them, you know, all the time, every day and when I see how incredibly sophisticated their quote unquote thinking is and I say quote unquote because on the other hand I'M not a member of that sort of hype group that says they're conscious and they can think and that, you know, they're going to take over the world or something like that. They have some very, very amazingly sophisticated behaviors and they can say things which sound as though they're introspecting on their own life, you know, well, I made this decision because I did, you know, I had this thought and so on and so forth. But whether or not that means that they're actually conscious and aware I think is a relatively open question. But I think at this point the answer is pretty much no, but not completely rule outable as they become more and more sophisticated. Right. You know, and there are a lot of reasons technically and otherwise to doubt that what you're seeing right now is genuine conscious behavior. I mean, you know, an analogy I think of is here in LA now we have a lot of those self driving waymos and they've only been active here I think for a little less than actually ChatGPT has been around so like, you know, maybe a year or two and longer. In San Francisco, I remember seeing them up there a couple years ago and right now you're really not seeing, you're seeing a lot of them on the road, relatively speaking, but you're not, it's, it's not like every car is a self driving car.

[Podcast Host]

Although it seems like it sometimes.

[Archever Blackwell, PhD]

Yeah, no, it does seem like it sometimes. And so I think that what, in the same way, it's just very early, I think someone's saying oh gosh, you know, there's not enough self driving cars on the road and so it's just never really going to take off. And let me think it's the same kind of thing.

[Podcast Host]

Let me take a turn on that though, if I may. Yeah, if you're, if you're advising business today, and I won't be more specific intentionally, how much do you would advise them that they should be building agents for almost every key function and, and have it set up so that agents can communicate with agents and then that productivity that gets assigned to them also has a metric associated with that productivity.

[Archever Blackwell, PhD]

Yeah, I think that's exactly what should be getting done. Now what I'd point out is that an analysis of exactly what goes wrong when some companies put AI into practice. One of the biggest problems that they found is that the implementation doesn't integrate into the existing workflow and the existing dynamics of the company. So they'll sort of, basically they'll bolt on some AI component in some way or another that's over here and then they have their workflow over here and they really haven't made it easy. So there's a lot of copying and pasting or there's a lot of, you know, I have to reformat this information in a way for the AI to accept it. And so people just don't find that to be a particularly useful utility for them. And so I think that's not the fault of the AI itself, that's the fault of the way it's been integrated in. And I think that as we get more savvy about how to integrate these things in and how to use these as part of our workflow, that problem's become going to become less and less prevalent and that's when you're going to begin to see some real gains in my view.

[Podcast Host]

It's interesting you mentioned the early days of computers and networking and yeah. Remember all those challenges we had with Token Ring and Netware and interoperability and that brought in the rise of consultants and then global integrators and you had all these people that were essentially a channel of distribution, became a channel of, of services that had to make sure everything worked and then they became sort of the top of the pyramid in terms of who was going to get it done and who was going to realize the productivity gains, who were the best at performing those services.

[Archever Blackwell, PhD]

Right. I remember when I was at Ask Jeeves many, many, many years ago.

[Podcast Host]

Thanks for mentioning that poll. That was a great one.

[Archever Blackwell, PhD]

Yeah. A name that you know, is probably not going to be familiar to everybody, but hopefully some people remember it.

[Podcast Host]

How about the logo with the Butler?

[Archever Blackwell, PhD]

Yeah, the, but the Butler logo. I remember that. I, I missed that. I, they, they now it's just like ask.com or something generic, but.

[Podcast Host]

Right.

[Archever Blackwell, PhD]

I remember, you know, that we were talking about how as soon as we got major funding, like one of the first things was just to write a huge check to Oracle to design the database. And that was kind of the default behavior at the time for any kind of startup. And of course now you could just use a cloud database that you know, say Google provides or. I mean all of the, a lot of those backend infrastructure problems have become, have essentially not gone away, but have become much less costly for startups to deal with. And I suspect you'll probably see that kind of trend continuing with AI as well because there's just such a huge amount of activity going on in that space and it's got such an amazing potential and what it's even Just if you see what it's doing now, not just potential, but what's, you know, what's going on.

[Podcast Host]

So when you look at, at this scenario, where do you see the big opportunity? I know this is part of your book, but just in general, where are you, where are you getting excited when you see the next 24 months evolve?

[Archever Blackwell, PhD]

Well, there are a lot of really cool opportunities. I mean, Anthropic is doing some very good work. They do Claude and Anthropic, for instance, has just released what they call Claude Cowork. And this actually can access the files on your hard drive and it can also sort of prod you to continue to work on whatever project. So it's not so much a passive recipient of what human beings are typing into it, it's much more of a really a co worker. And I think that that's the way. As you know, in the next 24 months what we're going to see is not someone dumping something into AI and then getting a response and posting it. We're going to see that the real value of AI is a cooperative system. A collaborator will work hand in glove. So I'll write up something and the AI will give notes and some feedback on how well it thinks it's going to hit and then I'll respond to that and it'll be sort of a virtuous cycle of each system. So the AI, eventually, I think what will happen is it will be able to learn your particular pattern of work behavior. So right now, frontier models, they're kind of siloed. So when you use it, there is a little bit of carryover. But basically each conversation is essentially a new conversation with the AI. And I think what you're going to see is first of all, hand in glove sort of cooperation, be some learning and carryover. And I think the other thing that isn't really so much on people's radar now, but that I think you're going to see is really most people, when they think of AI right now, they think of the frontier models, the chat GPTs, the Google Geminis, the, you know, clauds. But there's an enormous amount of potential in using local large language models that are just resident on your own machine and that don't necessarily have to connect to the Internet in order to do what they do. And these are smaller models, but I've worked with them and it's amazing how they're remarkably powerful for their size. And of course all of these models, the local ones and the frontier ones are all just going to get better and better and better. So I think you're going to begin to see people recognizing that local language models, local large language models have a lot of advantages because first of all, a lot of the use cases that businesses have are require confidentiality. So with medical, with digital agencies, which is what I talk about in my book, that digital agencies, their concern is, you know, we don't want our intellectual property to end up as training data for ChatGPT and then end up in somebody else's deck. You have the same thing with, you know, medical information, obviously, and with legal information. And so you can run a complete large language model locally without it ever being connected to the Internet and still get some very useful, powerful results from that.

[Podcast Host]

Save a lot of money too.

[Archever Blackwell, PhD]

What's that?

[Podcast Host]

You save a lot of money.

[Archever Blackwell, PhD]

You would say you would. Absolutely. You would save a lot of money. Because the thing is that the frontier models are very powerful. But how much of that power do you necessarily need for your particular use case? Yeah, I think you're making.

[Podcast Host]

Don't need it.

[Archever Blackwell, PhD]

Yeah, exactly. If you're looking for some kind of general Swiss army knife supercomputer style model, then something like ChatGPT or Claude is going to be the thing you're going to choose. But if you're looking for very, very specific kinds of use cases and they're very narrow, so you're talking about a digital agency that wants to get ideas and information and, you know, work content for a new toothpaste or something like that. Right. Then you really don't need all of the power of those large frontier models.

[Podcast Host]

You know, you bringing up this interesting point which I haven't dwelled on much, the localization of LLMs.

[Archever Blackwell, PhD]

Yes.

[Podcast Host]

I've heard we talking a lot today about the verticalization of lm, where a lot of the apps are more horizontal in nature.

[Archever Blackwell, PhD]

Yeah.

[Podcast Host]

And those are becoming very popular. But there's also a lot of push now towards medical AI.

[Archever Blackwell, PhD]

Exactly.

[Podcast Host]

And what's next? Plumbing AI or golf AI or real estate AI?

[Archever Blackwell, PhD]

Why not tune it? And that's the other thing that you have with large language models that are local is you can tune them on your own material and still keep them offline so that they are tuned. One of the things I talk about in my book is that the problem a lot of people have with using AI for creative work is everything kind of comes out sounding the same. And you want to preserve the voice of your agency, the voice of your product. When you're doing any kind of, you know, advertising outreach, whether It's a Facebook post or, you know, a print ad or what have you. And what you can do with large language models that are local is you can actually give it snippets of your previous work product. You can give it, you know, old ad campaigns and so forth, and have it learn to use that voice in anything that you create afterwards. So you train it on, you know, hey, this is like our past year's work of work, and we're a very hip Gen Z agency, and this is the kind of hip stuff that we provide. And then you say, okay, now I want to sell this new coffee house. And. And it writes it not in sort of generic AI form, but in something that's very specific to the voice that you've already established. And it works really well. It's really quite amazing that these relatively small models can be tuned so well. And the thing is, you can't do that with Frontier models. You can't tune them. What you do is, you know, you send it off to ChatGPT or Claude or Gemini and you get a response. You can do a little bit of tuning by playing with the prompt, of course, but it's not, generally speaking, going to be as powerful. And again, you're going to have the privacy concerns. So I think that's one of the things that you're going to see in the next 24 months becoming more prominent. And cost is a big factor too, as you point out.

[Podcast Host]

I also think it's kind of funny, as you point out your books, the Creative Agency Guide to AI. And most of the shots AI takes is it doesn't work for creative. Right. It's sort of not the thinking dialect you're looking for.

[Archever Blackwell, PhD]

And that's one thing that I address is because a lot of AI does come out with this very generic sort of homogeneous voice in so much of the creative work it provides. And my thesis is not that you give the AI something and it does it and it's gone. My thesis is that you can use it to generate ideas for you that you can then, you know, filter through and select to give you feedback and targeting. One thing that I've had a lot of experience with is Facebook targeting. Yeah. And what's really interesting to me is that previously with targeting, you had to basically just let the ad go out there on Facebook and see how people reacted and that there was, you know, a turnaround time to that. And if you weren't getting traction with it, then you'd have to tweak it and try to get something else. What AI allows you to do. Most people don't. Most people think of AI as creating content, but AI can also create. Give you some judgment on content. And so what you can have it do is say, hey, for this particular kind of content, this is who I think you're most likely to be successful targeting. And of these 10 targets, this is how I would phrase what you're trying to say in this document to best reach your target on Facebook or whatever your target is.

[Podcast Host]

That's really a good point. I'm glad you brought that up. Just, I sort of tease myself with experimenting in such ways, but really nail it. Where do you stand on the, on the development of AGI and the whole next quantum leap in technology?

[Archever Blackwell, PhD]

You know, I'm. I think that I'm pretty agnostic about that. I. I don't. I'm not a big fan of making predictions because usually they're wrong. I think that what's going to happen is that these systems will become more and more sophisticated and some people might interpret that as being, you know, AGI, and some people will say it's consciousness. The problem is that we don't really have a good understanding of how. What consciousness is in human beings. So, you know, we have a lot of theories. But as far as what's going on, is it because we're, we have carbon and neural tissue and these things are silicon, which is a different substrate? We really don't know. So, I mean, if we're talking about AGI, the sort of limit of it is this notion that it's actually going to become aware and alive and conscious and like, maybe, you know, start killing people like Hal in 2001. And I don't really think that that's what we're going to see happening. I think we're going to see amazing sophistication, but they're also going to continue to be somewhat passive. I mean, I think my view is probably the best example of what they're going to be like is the computers on Star Trek. So on Star Trek, they're incredibly sophisticated, but they're also essentially passive tools to be used by, you know, whoever the people, Captain Kirk and Picard and all those people. So the computers never kind of take over and they're not ever like characters in the story or anything, like, except for Data, I guess. And so I think it's going to be. You'll see incredible sophistication, but they'll still essentially be tools waiting for us to interact with them. They're not going to be like, hey, I'm here, I want to do this, I want to do that. I have agency until maybe, you know, I don't know what. There's a lot of theory around the notion of these models will really take off when they're embodied in some kind of a humanoid, you know, robotic, or not even humanoid, but some kind of robotic structure. The reason being they'll be able to learn continuously by interacting with the environment, and that will give them actual agency. And that's a very interesting idea. I mean, we already have humanoid robots kind of as prototypes. So I'm open to those, you know, becoming extremely sophisticated. And I always said I never studied because people were like, well, you're so interested in computers, why didn't you ever study robots? And I always say, you know, I don't want to study something that could come after me. And, you know, that's the thing is that they could like potentially, usually when you see them, they're kind of in these pens.

[Podcast Host]

Oh yeah.

[Archever Blackwell, PhD]

They don't go crazy within research environment. So they don't go crazy and start killing people. And whether that's killing people out of malice or just a malfunction doesn't kind of matter if you're the dead guy.

[Podcast Host]

Right? So I've seen those movies too. Yeah, exactly.

[Archever Blackwell, PhD]

We've, we've all seen.

[Podcast Host]

I want to, I want to steer this conversation coming around the bend here. Again, appreciate your time and wisdom. The idea that I reach a lot of early stage companies that are building AI tools, technologies, software, applications, programs, platforms, etc, and I, and I know you do as well, you come from an academic orientation and how do you, how do you coach in an environment like the one we're in today that changes so rapidly. So we've got the playbook from the computer era we sped forth. Now we have the playbook from the SAS era. There's a whole bunch of other industry segments, of course, that have their own playbooks. But now we're in AI, do we throw all that away and start anew or how do I go about building a company for this era, for this day and age? Really?

[Archever Blackwell, PhD]

That's a great question. I think that in terms of coaching new talent for those kinds of challenges, I think people are going to have to take much more responsibility for their own education and actually go be much more proactive about going out there. So, you know, this notion that you go to your local college or extension and you take a course or something like that, I think that's probably not going to be as feasible for people in the future because first of all, things are changing so fast that every year the course is going to have to change. Yeah. And I think that what you see now is people are doing much more of this coding through AI and because of that that's really kind of the skill they have to learn, which is not the coding per se so much, although you still should understand how code works, but rather the ways that you and how to optional optimally interact with AI so that it provides you with the output and the result, the work product that you're trying to get. And so I think that for myself I've seen it's much more of a sort of personal. You have to really be super motivated and be a self starter, I think.

[Podcast Host]

Yeah.

[Archever Blackwell, PhD]

And you have to just point. Yeah, what's that? Yeah, you have to just. I'm going to go in there, I'm going to interact with these systems on a regular basis. I'm going to do, you know, I do things like I'll have create a Colab project which is a Google, basically a sort of a little research tool that Google provides people for free. And so anyone can create Google Colab projects and use them to step through code and to look at what these systems are doing. And I think that understanding those aspects of AI is probably going to be much more important than understanding how to code a line of Python, although that's, you know, still going to be important. But we are seeing tremendous upheaval in, you know, the kinds of things that companies are looking for and a lot of them I think just aren't. Unfortunately a lot of them just know there's this thing called AI and they need to know about it, but they don't have really much of an understanding of it.

[Podcast Host]

Yeah.

[Archever Blackwell, PhD]

And I think the only thing that's going to solve that problem is time.

[Podcast Host]

It's going to take a lot of work.

[Archever Blackwell, PhD]

It's going to take time and a lot of work. And these things always, you know, if you look back to the Internet, I mean it was really many, many years from the beginning of the first web browser and the first website to when it just became part of the economic life of, you know, the world basically.

[Podcast Host]

So given, given your study and your focus, the new book on the creative agency, are you looking at a new business yourself? Do you have a new idea? What, what do you do with those great ideas?

[Archever Blackwell, PhD]

Yeah, I'm actually working on a couple of things. One of them is basically is a local LLM for creative agencies and you know, potentially for other kinds of.

[Podcast Host]

Oh, please keep US informed on that.

[Archever Blackwell, PhD]

Yeah, you can actually get a little more information about it at. Well there's my newsletter first which is insidetheblackbox AI.

[Podcast Sponsor Announcer]

Great.

[Archever Blackwell, PhD]

And then there's also yourvoicecraft AI which is a link to the book but also talks a little bit more about the application itself. And basically the application, you can think of it as like chat GPT for private local use. And so what happens is we give the client a local system that they can run if they wish to completely air gapped, not connected to the Internet in any way. And we get from them, you know as I was talking about before, we get from them basically snippets kind of like, you know, what a bloodhound would get to kind of learn what it is you're looking for. Those get the system gets trained on those which doesn't take very long and then they have a completely closed system that they can use in order to do in this first case, you know, digital agency work. And I've done a little of this, you know, testing it and so forth. Not a lot of it actually and it works very well. And eventually I'd like to see it sort of scaled to other domains where privacy and security are concerned. So I actually worked at a company that did legal document analysis using computers and linguists sort of combined. And again it was this sort of human and computer together model and the security requirements there were enormous. You know, I mean we had to actually black out windows where the, the legal material. Yeah, it was kind of crazy. The legal material was stored because they were afraid, you know, someone would try to peer in with a telescope. I, I don't know what they were afraid of but. So that's one of the, one of the, basically the major thing I'm working on right now, a couple other little projects but I, you know, I won't get into that but I think that the, the local language model is going to be for a lot of people where their buy in to AI starts with or at least is an important component of.

[Podcast Host]

You're fantastic. I'm really intrigued. I want to continue to follow up. The book is the Creative Agency Guide to AI.

[Archever Blackwell, PhD]

Yes.

[Podcast Host]

Thanks for sharing some time. You have any last parting thoughts that you want to share?

[Archever Blackwell, PhD]

No, I've loved this. This is great. I love talking about this. This is a topic I'm love to talk about.

[Podcast Host]

Well, I will drop. I make sure in our show notes we'll copy both the podcast, the blog and, and several other things because again this is, this is a story that just continues to unfold at a very fast pace.

[Archever Blackwell, PhD]

Yes.

[Podcast Host]

It's good to check in, so let's make sure.

[Archever Blackwell, PhD]

Very amazing. What's going on. Yes, Great. Thanks. Well, I've really enjoyed it.

[Podcast Host]

Thank you so much.