I think tool calling is your answer—you’re just missing a separation of concern(s).
For example to handle personality configuration, don’t use a tool to get personality configuration , use a tool to handle responding to the customer. When your agent has gathered the information to respond to the customer it will call the tool sendMessage with the response. Your tool call implementation is a role play prompt that rephrases the message with the provided tone/personality configuration (this is where the customer config is injected as context). The output is then passed through a guardrails completion for potential censoring before finally being displayed to the customer.
This means your main agent model simply becomes a routing agent (a tool calling optimized model) that directs to sub agents that handles various tasks (like figuring out shipping capabilities, or flavoring responses with personality affects, or adhering to guardrails) keeping the customer centric configuration’s blast radius (impact on the effectiveness of your prompts) narrowed to purely aesthetic completion and out of any functional completion.
FWIW, I think this article could just as accurately be titled "Diagrams Developers can, and cannot, generate".
I'm mainly speaking to the ability to read IaC code ([probably of any library but at LEAST in my case] cdk, pulumi, terraform, cloudformation, serverless) and be able to infer architectural flow from it. It's really not conducive to that use case.
I could also, kidding/not kidding, be speaking to the range of abilities for "mid" and "senior" developers to know and convey such flows in diagrams.
But really my point is this feels like more validation that AI doesn't provide increased ability, it provides existing (and demonstrated) ability faster with less formalized context. The "less formalized context" is what distinguishes it from programs/code.
I found the animation end point and start point to be rather abrupt. Can you add/tweak the easing function so the stopping isn’t sudden? For me it produces a cognitive thunk that takes me out of concentration on breathing.
I though about this movie a lot going through IVF with my wife. Particularly when selecting the embryo to implant. For those that don’t know:
With IVF, the doctor harvests as many eggs from the woman as possible. Then after taking the make sample and cleaning, selecting the semen with best modality , fertilize each embryo. Then let the embryo grow for a couple of days then freeze the embryo until it’s ready to be implanted (likely during the woman’s next cycle). If you choose, and generally for a fee, you can have each embryo tested for chromosome mutations.
After the process your doctor will call you and let you know how many eggs were harvested and how many were fertilized. They can then give you a report card for the embryos that represent likelihood of live birth, chromosome evaluation, male/female etc.
You then make a selection of which embryo (s) to insert.
Obviously, given the expense of IVF it’s hard to imagine a scenario of not telling the doctor to pick the highest rated (male or female) embryo just for best chance of success.
Felt very reminiscent of gattaca. And that was just selecting the best that we produced. I can easily see myself saying yes to “The best embryo you have had a chromosome defect that will likely result in condition X, but we can fix the defect with a small gene modification, and give the child the best chance for a healthy life, should we modify the embryo?”
And then the slippery slope, “we can also improve the odds of higher intellect, being taller , thinner, etc”.
Did you find a very skewed sex ratio in your embryos?
My dad has a sister, but has only two sons. My brother has 3 sons and one daughter, all natural pregnancies. My wife and I have 1 son via IVF, and the other 4 embryos we had checked for chromosomal abnormalities happened to all be male.
I am aware that some IVF processes can inadvertently introduce a sex bias by exacerbating the tiny speed advantage the lighter Y chromosome gives sperm.
I'm curious (and will never know) how much of the skewed sex ratio of my embryos is due to randomness, how much is due to IVF processes, and how much might be due to some familial abnormality.
We did. The majority were female embryos. I mentioned report cards, we had 5 embryo's with grade A, 1 of which was male. We had a similar prevalence for female with lower grades as well, though there were 2 or 3 males each in the lower grades.
There are some harmful genetic mutations that associate with particular sex chromosomes. I wonder if ease of detection in some varieties of abnormality may play an inadvertent role too.
I'm always confused. The NWO conspiracy people say that gene editing will be prohibitively expensive (like owning a yacht is), but the real life examples I see don't seem that way. What are some good examples of services that could be done cheaply but are intentionally made expensive for the sake of exclusion?
> but the real life examples I see don't seem that way.
In what situation would removing an egg, fertilizing it, altering its genetic sequence, and them implanting the embryo into a person be consider cheap?
Looks like IVF costs around $12,000 today. Considering that this is for your own child, I bet it would be one of the least expensive parts of raising them. I also expect it would become very cheap after it went mainstream.
> The Human Genome Project was the international research effort to determine the DNA sequence of the entire human genome. It took 13 years and was published in 2003, with an estimated cost of over $300 million. Today, a whole human genome can be sequenced in one day for under $1000.
$12,000 is more than the average yearly income for a person on this planet. The USA has only 300 million or so people in it. What are the chances you are born in a place where $12,000 is considered 'the least expensive parts of raising a child'?
A mobile phone can be mass produced based on one design. This is not the case with removing eggs and doing custom gene work on them, incubating them, and implanting them in a person. You are making an invalid comparison. There are no 'economies of scale' with individual procedures.
One interesting balancing thing though is that because people who do IVF tend to be older / have more high risk pregnancies, you’re more likely to have issues there.
Amazing! Love it so much. My brain was running wild with possibilities (like having an autocomplete from the corpus, live audio only previews, and the above)
I didn't realise there was a github! You should add a link to your tutorial
I guess there would need to be an intermediate step. Videogrep helps to surface useful ngrams and there would still be a manual/creative step to stitch them together in a way that works.
IMO the next Google needs to use AI to classify results as fact/fiction and supposition/exposition etc. It will identify bias and even classify it (religious, political, cultural etc). With the abundant sources of information (and misinformation) ever growing, the job of a search engine will be to inform you about the information you are viewing and alert you to key factors that may be coloring the information.
I like the idea of a repairable/upgradable/modular laptop. However, to really buy in to the idea I want more than a promise of future upgradability. I'd really like to see a company roadmap that shows expected future dates of upgrade releases.
Show me if you're expecting to put out new CPU upgrade parts every 1, 2, 3 or 5 years.
Show me what type/generation of graphic card is available and your expectation of how far behind graphic card modules will lag behind current gen cards, 1,3,5,10 years?
Show me how long I'll expect to have to wait to double my storage, or ram.
And most of all, what are the target price points of current and future upgrades.
roadmaps are meaningless. talk is cheap. Just ask anyone who has ever bought into a "live service" video game, or countless other ambitious but later abandoned products.
Here's what I can say for sure. The options for future upgrades will be correlated with the sales figures of the base laptop.
> My idea for an 11-star experience 1 in finding new books is that Goodreads knows me even better than I know myself and constantly recommends the perfect book.
> goodreads shows me five books that I don’t want to read.
I wonder if these two ideas are at odds with each other. Imagine for a moment that recommendation engines were solved problems, and definitely worked given the above statement. They know you better than you know yourself. Would it’s recommendations likely only include books that you obviously wanted to read? Or would they include books you didn’t know you wanted or needed to read? I mean this in terms of judging the books by their cover rather knowing about their existence. Isn’t it likely or even probable that the majority of books recommended would be based on the value they contribute to something deeper than the pure enjoyment purposes?
As an aside, I remember in a college literature class I took the instructor told us that it’s up to the reader to derive value or meaning from stories. This was a class that studied short stories of early American authors. Most of which were slice of life narratives that didn’t have any apparent meaning, or commentary from the authors themselves. The exercise was to study the characters, scenery and tone and try derive what might either lie beneath the words or story themselves. Whether the ideas we deduced from the stories were accurate (and in most cases probably were not) the value Of the process was of critical thinking about the stories that made us consider and express ideas and beliefs we normally don’t.
Back on topic, would a working recommendation engine likely suggest things that on the surface seemed either boring or blatantly unappealing that would provide tremendous value if we put work in to reading and studying?
That being said, is it possible that current recommendation engines are already working? Most are at least driven by reading behaviors of the masses, so it seems like it might be feasible that that Steven king book that is unappealing to you is something you should actually read.
(This is not including recommendations for books you have already read in different languages which seems like an obvious bug, but then again reading books you’ve already read in different languages might be an excellent way to become more fluent in a new language or gain a deeper understanding of translation of ideas between languages....)
So, maybe the tech isn’t something that needs to be fixed. Maybe we just need to be open to what the tech is telling us?
I think you are right. Being recommended a book I've never heard about by some author I'v never heard of by some computer system, I'm probably gonna dismiss it. However, if I've seen it in a bookstore, read about it in a paper, heard a friend talk about it etc. then I might act on the recommendation.
Basically being exposed to something enough times. First time and a cursory look, most things don't look to exciting. This goes for everything. Movies, restaurants, gadgets..
Seems to me like SSN is missing an open free (as in paid for by taxes) 2FA service.
When I obtain a drivers license or passport (or some other process to confirm my identity at a government facility) that I should be able specify my 2FA medium (email address or phone number [text or call]).
Any organization that wants to prevent identity theft should be able to do a 2FA challenge. As an individual I would receive a phone call or text or email "Acme Co. is processing a request to open up a new line of credit under your SSN. Did you initiate this action?"
There should then be regulatory requirements on banks, insurance, etc that require 2FA confirmations before associating an SSN to your account.
This means your main agent model simply becomes a routing agent (a tool calling optimized model) that directs to sub agents that handles various tasks (like figuring out shipping capabilities, or flavoring responses with personality affects, or adhering to guardrails) keeping the customer centric configuration’s blast radius (impact on the effectiveness of your prompts) narrowed to purely aesthetic completion and out of any functional completion.