Remodeling Generative AI Into Area-Savviness By way of In-Context Studying And A Sprint Of Knowledge Engineering Appears Promising, Says AI Ethics And AI Legislation

Which is healthier, being a jack-of-all-trades or being knee-deep in a specific area or specialty?

You’d be hard-pressed to say that one or the opposite is essentially higher or finest. All of it depends upon the circumstance at hand.

The jack-of-all-trades is presumably acquainted with quite a lot of subjects and is aware of a tad about every subject. In the event you have been misplaced in a forest and had a jack-of-all-trades with you, the individual may be vaguely conscious of general survival strategies and will present wilderness steerage accordingly.

Suppose although that whereas within the forest, you fell and broke your leg and cracked your ribs. The jack-of-all-trades ally would possibly know a semblance of first assist akin to placing a crude splint in your fractured leg. In the event you had a medical physician with you relatively than a jack-of-all-trades, and also you perchance suffered the identical accidents, the percentages are that the medical professional would be capable to undertake extra finely and deeply wanted medical care.

I carry up the traditional query of breadth versus depth to showcase how the identical consideration is taking part in out in the present day in regards to the introduction of generative AI or giant language fashions (LLMs) such because the extensively and wildly common ChatGPT by AI maker OpenAI, and others akin to AI apps by Google (Bard), Anthropic (Claude), and so forth. We’re mightily confronted with a momentous and unresolved dilemma of whether or not it’s higher to knowledge prepare and devise generative AI on a generic jack-of-all-trades foundation or a extremely specialised foundation in a specific area of curiosity.

In in the present day’s column, I shall be inspecting this relatively imposing and vexing challenge that confronts anybody wanting to make use of generative AI for domain-specific inquiries. The problem is that almost all generative AI is devised on a totally generic foundation and might be construed as a jack-of-all-trades. You aren’t doubtless capable of keep on any wise in-depth dialogues in say medication, authorized, and different explicit domains with most generative AI since that’s not how the generative AI was set as much as start with. The prevailing method is to plan generic generative AI that’s primarily an professional on nothing and a chatterbox on absolutely anything.

This doesn’t appear to be stopping individuals from utilizing generative AI in ways in which they shouldn’t.

I not too long ago coated for instance the information story concerning the two attorneys that relied on ChatGPT for doing their authorized analysis in a real-life court docket case, see the hyperlink right here. OpenAI notably warns that ChatGPT shouldn’t be natively appropriate for such a job (for my protection of the thought of out-of-scope or prohibited makes use of of ChatGPT, see my dialogue at the hyperlink right here). The attorneys ended up making use of fictitious or so-called AI-hallucinated authorized precedents and acquired caught by the opposing aspect and the decide for offering falsities to the court docket. That’s a no-no.

Equally if not scarier is that at instances there are medical professionals that appear to be relying upon generic generative AI, as I’ve famous at the hyperlink right here. This can be very simple to be lulled into believing no matter a generative AI app tells you, particularly if the content material appears believable. Moreover, in case you repeatedly use generative AI and it “all the time” appears to be right, you begin to assume that it’ll henceforth be completely dependable and apt. Any of us can fall into that psychological entice.

The gist is that generative AI shouldn’t be often formed towards explicit domains of experience. As a substitute, generative AI is typically knowledge skilled throughout a swath of content material that covers almost any subject beneath the solar. The standard technique consists of information scanning throughout hundreds of Web web sites and utilizing the discovered content material to knowledge prepare the AI. That is to an ideal extent why the generative AI seems to be fluent general. The purpose is to cowl our use of pure language in a broad sense.

Let’s take a second to ensure we’re all on the identical web page concerning the foundational keystones of generative AI. After doing so, I’ll dive additional into this query of breadth versus depth.

Keystones About Generative AI

Generative AI is the newest and hottest type of AI and has caught our collective rapt consideration for being seemingly fluent in endeavor on-line interactive dialoguing and producing essays that seem like composed by the human hand. In short, generative AI makes use of advanced mathematical and computational pattern-matching that may mimic human compositions by having been data-trained on the textual content and different content material discovered on the Web. For my detailed elaboration on how this works see the hyperlink right here.

The standard method to utilizing ChatGPT or another related generative AI akin to Bard (Google), Claude (Anthropic), and so forth. is to have interaction in an interactive dialogue or dialog with the AI. Doing so is admittedly a bit wonderful and at instances startling on the seemingly fluent nature of these AI-fostered discussions that may happen. The response by many individuals is that certainly this may be a sign that in the present day’s AI is reaching a degree of sentience.

On a significant sidebar, please know that in the present day’s generative AI and certainly no different sort of AI is presently sentient. I point out this as a result of there’s a slew of blaring headlines that proclaim AI as being sentient or at the very least on the verge of being so. That is simply not true. The generative AI of in the present day, which admittedly appears startling able to generative essays and interactive dialogues as if by the hand of a human, are all utilizing computational and mathematical means. No sentience lurks inside.

There are quite a few general issues about generative AI.

For instance, you may be conscious that generative AI can produce outputs that include errors, have biases, include falsehoods, incur glitches, and concoct seemingly plausible but completely fictitious info (this latter aspect is termed as AI hallucinations, which is one other awful and deceptive naming that anthropomorphizes AI, see my elaboration at the hyperlink right here). An individual utilizing generative AI might be fooled into believing generative AI as a result of aura of competence and confidence that comes throughout in how the essays or interactions are worded. The underside line is that you should all the time be in your guard and have a relentless mindfulness of being uncertain of what’s being outputted. Be certain to double-check something that generative AI emits. Finest to be protected than sorry, as they are saying.

Into all of this comes a plethora of AI Ethics and AI Legislation issues.

There are ongoing efforts to imbue Moral AI rules into the event and fielding of AI apps. A rising contingent of involved and erstwhile AI ethicists are attempting to make sure that efforts to plan and undertake AI takes into consideration a view of doing AI For Good and averting AI For Unhealthy. Likewise, there are proposed new AI legal guidelines which might be being bandied round as potential options to maintain AI endeavors from going amok on human rights and the like. For my ongoing protection of AI Ethics and AI Legislation, see the hyperlink right here and the hyperlink right here.

The event and promulgation of Moral AI precepts are being pursued to hopefully stop society from falling right into a myriad of AI-inducing traps. For my protection of the UN AI Ethics rules as devised and supported by almost 200 international locations through the efforts of UNESCO, see the hyperlink right here. In the same vein, new AI legal guidelines are being explored to attempt to maintain AI on a good keel. One of many newest takes consists of a set of proposed AI Invoice of Rights that the U.S. White Home not too long ago launched to establish human rights in an age of AI, see the hyperlink right here. It takes a village to maintain AI and AI builders on a rightful path and deter the purposeful or unintended underhanded efforts that may undercut society.

That fast rundown of what’s up with generative AI ought to hopefully put us all in the identical mindset and permit me to subsequent dive into the breadth versus depth conundrum.

Making A Dive Into The Breadth Versus Depth Conundrum

You may be tempted to assume that the breadth versus depth conundrum underlying generative AI is definitely resolved. It isn’t.

Contemplate these 4 distinct potentialities:

• (1) Breadth-only generative AI. Devise generative AI that’s the proverbial jack-of-all-trades, which is the predominant method in the present day. That is what you often use.
• (2) Depth-only generative AI. Craft generative AI for a specific area such because the medical or authorized specialties, however this has potential downsides that I describe subsequent. That is sometimes undertaken.
• (3) Construct generative AI as each mixed into one. The thought right here is to attempt to infuse each breadth and depth into one generative AI, which is an experimental and considerably controversial method presently being pursued. I’ll say extra about this under.
• (4) Momentarily garner each as if blended into one. That is the newest method whereby you often begin with a breadth-only generative AI after which do some intelligent trickery to often briefly get a depth-added taste happening.

You already know what a breath-oriented generative AI is (my first listed bullet level above).

That’s what you’re doubtless to make use of while you make use of an off-the-shelf generative AI app these days. No want for me to say rather more about it, apart from to level out that you just shouldn’t be wantonly utilizing breadth-only for making an attempt to do depth-oriented interactions. You are attempting to place a spherical peg in a sq. gap. Cease doing this except you’re doing so with appropriate issues at hand (I’ll cowl this shortly).

Let’s study subsequent the state of affairs of aiming to develop or devise a depth-only generative AI app (my second bulleted level of above).

Think about that we started with a clean slate of a generative AI. There isn’t any content material in any respect. We merely have a generic pattern-matching engine that may grind via no matter knowledge we decide to scan. For the second, this generative AI is at zero since you’ll be able to’t use it for something in any respect apart from to first knowledge prepare it.

Suppose we wish to flip this empty shell right into a generative AI dedicated to the medical area.

Right here’s the deal.

We handle to gather collectively all method of on-line and digital paperwork about medication. That would be the concentrated focus of the info coaching. There gained’t be something on this collected content material that covers all of the myriad different subjects of our trendy society. No knowledge coaching on social, political, philosophical, and different such content material. Solely medical stuff.

Would we get out of this a fluently interactive generative AI that was well-versed within the medical subject?

Possibly, however we are also probably undercutting the pattern-matching towards being doubtless underdeveloped in an overarching pure language all instructed. We’re omitting all of that different content material that covers day-to-day discussions and the way we write and compose our ideas. The percentages are that with out all of that different richness of content material, you might need a relatively stilted and marginally fluent generative AI.

We’d additionally significantly query whether or not this medically targeted generative AI goes to be as much as par in masking the human aspect of medical apply. To some extent, now we have stripped out humanness by solely utilizing simply medical literature and medical analysis. I’m not asserting that the generative AI gained’t perform. It most likely would. I’m simply noting that the end result might be going to be overly one-sided and never completely passable or totally sturdy.

Okay, that takes us to a want to mix or meld collectively each a breadth and depth set of approaches.

How would possibly you accomplish that?

Possibly you can do the breadth-first after which add the depth. Or, if that doesn’t appear finest, do the depth first after which add the breadth. One perspective is that it must not matter as to the order or sequence. If we’re making ourselves a scrumptious breakfast consisting of fried eggs and ham, presumably you’ll be able to put the eggs within the pan first after which toss within the ham or accomplish that the opposite manner round. The result’s hopefully going to be the identical.

Not everybody buys into that analogy in terms of cooking up generative AI.

One main argument goes that you just most likely must have breadth initially established so as to then dovetail or construct additional towards the depth. Some passionately argue you can’t as readily go in the wrong way. They might say that in case you select to first do the depth, akin to a medical or authorized area, after which try breadth placement on prime of this, the possibilities are it isn’t going to combine properly. The declare is that you should first set up the sphere or basis of breadth, and craft additional on prime of that, or else it would all get mired and tousled.

These are all fairly fascinating open-ended questions and varied makes an attempt and analysis investigations are underway to see to what diploma these competing angles are both promising or limiting.

We now come across my fourth listed bullet above, specifically the notion of briefly or momentarily making an attempt to get a mixing of breadth and depth. The excellence is that we aren’t going to totally construct the 2 into one. In lieu of that development, we’ll depend on having a breadth basis and see if we will smush a little bit of depth into the generative AI.

This smushing or melding will doubtless be on a short lived relatively than a everlasting foundation. The purpose is to get a generic generative AI for the second which is able to convincingly appear to have a gush of depth that we will leverage (I’ll cowl the upsides and drawbacks of this technique).

Brute Drive Is Not Going To Lower It

Let’s begin by contemplating the brute drive mindset or let’s consider ill-informed or harmless perspective some customers of generative AI appear to have.

An inclination by many on a regular basis customers of generative AI is to imagine that they will merely feed a ton of information into generative AI each time they want to get the generative AI to preach on a subject in a selected area. I name this a brute drive method as a result of the idea is that you just compile a big corpus of content material and purpose to jam it down the throat of your favored generative AI app. If one of many generative AI apps gained’t appear to take it, you’ll maintain making an attempt others till you discover one which appears accepting of the large corpus.

For instance, suppose you need your most well-liked generative AI app akin to ChatGPT or Bard to out of the blue be conversant concerning the intricacies of a uncommon species of North American hen such because the golden-cheeked warbler. The generative AI of its common generic capability doubtless doesn’t have many detailed sides about this uncommon hen. Knowledge relating to the golden-cheeked warbler most likely wasn’t particularly cultivated or scanned throughout the general knowledge coaching throughout the Web.

You determine due to this fact to take issues into your individual palms.

By accumulating collectively heaps and plenty of digital supplies concerning the warbler, you piece collectively a plethora of information and paperwork containing the intimate particulars of this uncommon hen. All you would appear to want to do is drive this down the gullet of the generative AI app. Ergo, you gleefully open up a immediate and attempt to feed all of this huge content material into the AI app.

Oopsie, you hit a limitation.

Few individuals understand that almost all generative AI has a comparatively constrained limitation on the dimensions of the context window when utilizing AI. I’ve mentioned at size why these limits exist, together with the time and price issues confronting the AI makers about making an attempt to enlarge these context home windows, see my evaluation at the hyperlink right here. It’s the avowed purpose of most AI makers to increase the allowed context window. That is thought of an vital technological consideration and will vastly enhance and make generative AI much more engaging to be used in lots of essential methods.

Backside-line presently is that you just can’t simply toss all the things together with the kitchen sink into your entered prompts. The generative AI will balk and inform you that you’ve got exceeded the allowed restrict. Worse, some generative AI apps don’t provide you with a warning that you’ve got exceeded the restrict. That is disconcerting because you would possibly merrily proceed beneath the completely false perception that your entire golden-cheeked warbler knowledge has been obtained and processed by the pattern-matching of the generative AI.

All proper, we can’t get a ten-pound field of birdfeed right into a container that can solely settle for two kilos.

What are we to do?

An intriguing method consists of well utilizing knowledge engineering precepts to set off the generative AI into doing in-context studying, appearing inside the context window constraints that you’re confronted with. This trick might be advantageous. Apart from probably getting the generic generative AI to turn out to be momentarily seemingly versed in a specific subtopic or subdomain, you’ll be able to proceed to make use of this method even as soon as the context window constraints are later sometime widened.

Permit me to elucidate that final level. The percentages are that the generative AI apps will more and more be superior to permit for bigger and bigger context window sizes. Will or not it’s giant sufficient for no matter you wish to do? In all probability not. Every improve in context measurement is undoubtedly going to be met by those who need the dimensions to be enlarged extra so. Alongside this staggered stepping course of, you’ll be able to maintain utilizing this explicit method and ratchet up accordingly to no matter context measurement window is offered on the time.

The Massive Mix By way of Utilizing Knowledge Engineering And In-Context Mannequin Studying

To dig into this ingenious method, let’s make use of a latest submit by AI researchers that describe their efforts to offer a kind of reference structure for doing this very form of momentary mixing. I’ll be citing a not too long ago posted paper entitled “Rising Architectures for LLM Functions” by Matt Bornstein and Rajko Radovanovic was posted on June 20, 2023. That is work beneath the auspices of the famend Enterprise Capital (VC) agency Andreessen Horowitz, which is often known as a16z or as AH Capital Administration, LLC.

Listed here are some key excerpts by the AI researchers:

• “The core concept of in-context studying is to make use of LLMs off the shelf (i.e., with none fine-tuning), then management their conduct via intelligent prompting and conditioning on personal ‘contextual’ knowledge.”
• “For instance, say you’re constructing a chatbot to reply questions on a set of authorized paperwork. Taking a naive method, you can paste all of the paperwork right into a ChatGPT or GPT-4 immediate, then ask a query about them on the finish. This may increasingly work for very small datasets, however it doesn’t scale. The most important GPT-4 mannequin can solely course of ~50 pages of enter textual content, and efficiency (measured by inference time and accuracy) degrades badly as you method this restrict, referred to as a context window.”
• “In-context studying solves this downside with a intelligent trick: as a substitute of sending all of the paperwork with every LLM immediate, it sends solely a handful of probably the most related paperwork. And probably the most related paperwork are decided with the assistance of . . . you guessed it . . . LLMs.”

As famous in these excerpts and per my aforementioned indications, you aren’t going to get very far by making an attempt a brute drive technique with in the present day’s generative AI. The context window goes to abruptly slap you down.

An alternate consists of exploiting or leveraging in-context associated “studying” related to up to date generative AI. In essence, you may get the generative AI to do its pattern-matching on a particular set of information that you just deliberately curate and mindfully feed into the AI app. You do that in a bite-at-a-time method.

The outdated adage to chew your meal as you eat applies to this method.

And, just remember to don’t chunk off greater than you’ll be able to chew.

Three Levels Of The Augmented Or Supplemented Generative AI

The AI researchers counsel that these three phases be used:

• “Knowledge preprocessing/embedding: This stage entails storing personal knowledge (authorized paperwork, in our instance) to be retrieved later. Usually, the paperwork are damaged into chunks, handed via an embedding mannequin, then saved in a specialised database referred to as a vector database.”
• “Immediate development/retrieval: When a person submits a question (a authorized query, on this case), the appliance constructs a collection of prompts to undergo the language mannequin. A compiled immediate sometimes combines a immediate template hard-coded by the developer; examples of legitimate outputs referred to as few-shot examples; any crucial info retrieved from exterior APIs; and a set of related paperwork retrieved from the vector database.”
• “Immediate execution/inference: As soon as the prompts have been compiled, they’re submitted to a pre-trained LLM for inference—together with each proprietary mannequin APIs and open-source or self-trained fashions. Some builders additionally add operational techniques like logging, caching, and validation at this stage.”

I’ll clarify what these three phases accomplish.

Within the first stage, you wish to acquire collectively no matter salient digital supplies you imagine are going to be wanted for the subdomain or subtopic of curiosity. You wish to be comparatively certain you will have a ample quantity and high quality of content material. Overshooting is usually going to be okay. Undershooting is much less helpful because you gained’t have sufficient content material or have insufficient content material that’s due to this fact unable to make sure the pattern-matching is carried out amply.

The collected digital knowledge is taken into account in uncooked textual content kind and will presumably be fed as bizarre text-based prompts. The factor is, we will and may decide to remodel that textual content into one thing extra readily useable by the generative AI app. This can assist velocity up issues while you wish to leverage the content material. You could preprocess the collected textual content and switch it into what is called embeddings, primarily a numeric internalization that’s pertinent to the specifics of the generative AI mechanisms you’re utilizing.

After doing the preprocessing, you’d then retailer the embeddings in a database specifically arrange for this particular form of utilization. The frequent parlance is to confer with this database as a vector database. The final idea is that the numeric transformation is in a collection of numeric vectors, nearly akin to what you realized about in these arduous algebra courses at college.

There are a selection of vector database administration techniques (v-DBMS) accessible and quite a few instruments for coping with vector databases. The AI researchers opted to say a couple of, such because the open-source Weaviate, Vespa, Qdrant, and others, together with describing vector administration libraries akin to Chroma and Faiss. This market phase is quickly evolving and increasing.

In recap, the primary stage has supplied us with a pre-processed collection of pattern-analyzed chunks throughout an array of significant components underlying no matter subtopic or subdomain is being pursued. When it comes time to make use of this, we will merely attain into and extract out of the vector database no matter parts we would have liked for feeding into the generative AI.

That was the primary stage.

For the second stage, we’ll fake that you’re now able to enter a immediate that shall be dependent upon utilizing the assorted reworked and readied materials present in your handy-dandy vector database that you just composed.

Your immediate might want to confer with the vector database that you just’ve compiled. The percentages are that you just’ll wish to depend on a templated immediate, one which already has the included connections over to the vector database. You’d sometimes be unwise to laboriously compose the wanted immediate from scratch, nor would you wish to accomplish that. As a substitute, you must prudently leverage a predetermined scheme that can do a number of issues without delay for you, akin to referring to the vector database, probably calling APIs (utility programming interfaces), and doing different under-the-hood mechanizations to make this work.

The immediate may also want to incorporate a number of few-shot examples. That is what will get the juices, because it have been, of the generative AI underway. The in-context studying is being stoked through the examples that you just present. The examples would assuredly must cowl no matter focus you bear in mind for the subtopic or subdomain you’re exploring.

That’s the second stage.

The third stage consists of submitting the immediate to the generative AI and permitting the entire equipment and kaboodle to be processed by the AI app. You’d then work together with the generative AI now that it has turn out to be considerably contextually fueled on the subtopic. You may even have the generative AI reply to questions and generate solutions or subdomain-focused essays for you.

The key sauce is that the generative AI is relying upon in-context oriented “studying” or pattern-matching related to no matter context you will have now supplied through your pre-processing and vectorized database snippets.

How a lot of this sticks with the generative AI is drastically dependent upon how the generative AI has been initially devised. It might be that just about nothing of the momentary domain-specific infusion will stay previous the top of the engaged dialog. You’ll primarily begin from scratch every time you enter a brand new immediate, although when you’ve acquired the vector database and the templated prompts in place, you don’t have to repeatedly reinvent the wheel on this. There may be additionally an opportunity that a number of the area specifics will gravitate into and turn out to be half and parcel of the generative AI (your mileage might range).

Surrounding Generative AI With The Added Items

The AI researchers acknowledge that a wide range of added instruments and elements are required to get generative AI to work on this augmented method.

They confer with an rising generative AI or LLM purposes stack, consisting of components akin to:

• Knowledge Pipelines
• Rising Fashions
• Vector Databases
• Immediate Few-Shot Examples
• Playground
• Orchestration
• APIs/Plugins
• App Internet hosting
• LLM Cache
• Logging/LLMops
• Validation
• And so forth.

Once more, these are all newly rising utility realms of immense market curiosity and development since they tie to and have the potential of notably enhancing generative AI as we all know it in the present day. These are elements that may assist flip a generative AI right into a extra versatile semi-specialized generative AI, albeit with important caveats and constraints.

Conclusion

Let’s agree that this method of utilizing knowledge engineering and in-context modeling shouldn’t be a silver bullet per se. It’s a helpful method that gives some key advantages. You’ll be able to to some reasoned diploma garner the upsides of a generic generative AI whereas having it additionally turn out to be partially steeped within the depths of a selected area.

One qualm is that the piecemeal dividing up of a wealthy area to leverage generative AI on this style is at instances nonetheless going to overlook the mark. You may not perchance ensnare the precise a part of the area and thus the generative AI is not going to be particularly conscious of your particular inquiries. Worse, and distressingly just like my earlier remarks, the generative AI may appear to offer totally responsive domain-specific responses and but be completely afield of what’s thought of right or correct in that area.

In that sense, some fear that this will make issues worse relatively than higher off.

Right here’s the rub.

If a generic generative AI can at the very least be detected at face worth as being off base concerning the depths of a specific area, think about how a lot trickier this may be to discern when the generative AI has been armed with the terminology and phrasing that appears fully aligned with that area. An individual utilizing generative AI may be extra simply bamboozled.

One other concern is the alluring temptation to “dumb down” the individual or individuals coming into the prompts pertaining to the area of curiosity.

Let me elaborate.

Suppose that we usually used a medical physician to enter prompts right into a generative AI that has been supplemented or augmented with this in-context mannequin extension in a medical area. We’d assume that the versed medical professional would use refined medical terminology and language pertaining to the medical area of curiosity. As well as, when interacting with the in-context domain-augmented generative AI or reviewing any essays produced, this identical medical physician would know what they entered and what they acquired again in return from the AI app.

A hospital decides that their medical docs are manner too costly for use for the mere entry of prompts into generative AI. As a substitute, this can be a clerk’s job, or so contends the hospital directors. Clerks are instructed that they’re to do the prompts on behalf of the medical workers. Simple-peasy.

Sadly, the clerks aren’t versed within the medical subject. The prompts they find yourself coming into are inflicting the generative AI to veer from what would have occurred had the prompts have been entered by the medical docs. Possibly no one catches onto this dire drift. In the meantime, the responses or essays being produced by the generative AI are being generated and floated round to the medical groups, which don’t actually know what led to the seemingly medically stout responses.

This might be a multitude. Not simply any mess, a multitude that additionally contains potential life and loss of life ramifications.

Maybe this helps emphasize the significance of AI Ethics and AI Legislation when considering how we’re utilizing generative AI and ways in which we’re augmenting generative AI.

A remaining comment for now.

In case you are interested in what a golden-cheeked warbler appears like, you’ll be able to go browsing and see fairly visually gorgeous photos of this velveteen-looking hen. Be certain to additionally discover a web site that incorporates the sounds of this treasured and uncommon songbird. I guarantee you these warblers make delightfully warbly sounds.

William Blake, the well-known English poet, stated that no hen soars too excessive in the event that they soar with its personal wings.

By augmenting generative AI to attempt to attain steeped domains, some ponder whether we’re making the leap in the precise methods. The generative AI is getting used to soar above its means, some are apt to sternly warning. Maybe we must put aside what appears to be an interim answer. Wait till we good a real seamless intermixing of each breadth and depth capabilities.

Not so, comes the retort. Any port in a storm. We are able to have our cake and eat it too.

Whichever manner this lands, be sure to contemplate whether or not you wish to make use of those augmented generative AI apps, and in that case, be sure your eyes are large open and you retain your thoughts, eyes, and ears open always. This isn’t simply hen watching, it might be life or loss of life.