What Siri should be = Inflection Pi

I predicted a while back, maybe a year ago that the whole chatGPT LLM (Large Language Model) "thing" will hit it's peak around Aug '23/Sept '23 and then decline towards Dec '23/Jan '24 where people will start looking at non-pay / non-monetised usage of LLMs.

I've also been a keen advocate of moving the usage / runtimes OFF of "other people's servers", ie. what you call "Cloud", because they are incentivised to implement vendor lock-in in subtle ways such as getting you to use a service that only they offer, or store your core data in a datastore that you cannot export / lift&shift elsewhere without it costing more than it is worth, therefore stealth lock-in.

I cannot really complain, businesses are in the business of business, therefore, they are driven by financial transactions and you, as the customer (still makes me chuckle that the "IT people" call customers "end users", just like drug dealers refer to their dependent locked-in customers/users), therefore they are not doing anything "wrong", but as a decent human being with morals, ethics & integrity you have to step back and look at the behaviour of some of these companies and decide if they really are "there to help you".

As I've said before, I've been investigating, since about Feb '23 running LLMs "offline", ie. on hardware such as your laptop or phone.  It was challenging at the start, I did "blow up" a massive spec laptop (an HP ZBook G7 - £8500!) and had to rebuild it a few times and as the tools / technology has moved on, so have I.  I now have ollama running on RPi5 devices just using CPUs and it's amazing.

Last year, I moved into the early usage of GPT4All, privateGPT, then langchain (I loved the release of this framework - this opened up the world to real usage LLM applications) and most recently into the usage of ollama (offline LLLM Engine) and the usage of Open-WebUI as a chatGPT4 style UI that allows you to point to many different LLM backends, either offline or online.

You will notice, all of the above are open-source.  ie. not monetised (yet), if you are in the area of "investigating" this technology, rather than trying to "sell" or force customers to adopt this technology in a production environment before it is ready beyond R&D then go away, stop reading, sod off - I'm not anti-LLM, I'm just pragmatic and logical / sensible and have removed emotion from my assessments.

Also, notice I do NOT refer to this technology as "AI", because it is not.  We do not have "AI" yet & I doubt we will have true "AI" in our life-time.  The "A" that most people refer to is "Artificial" - I would contest that the "A" stands for "Augmented" - ie. it is a tool that assists / augments you with a task... a bit like a buddy-by-your-side, someone to quiz & talk to at 3am when you cannot rely on a Google search or speaking to another human being for fear of ridicule or judgement.

BTW - you only need mega-compute on Cloud if your are "training" LLM models, which 90% of people will NOT be doing, therefore you do not need to use the Cloud, you can actually just use your laptop, or phone... do not believe the hype, think for yourself.

Right, let's get back to the point.  INFLECTION

I won't attempt to repeat what they say, in fact, I won't even write a prompt to summarise - read it as they wrote it:

What does that mean? Because I know you are a lazy f**ker and that text was small and it was an image and you are a busy person, so you don't have time to focus for 3 minutes to read the above - c'mon, be honest, I'm right.....be honest.... you skimmed it, didn't you.

Right, as the EPs (Early Professionals ) say to me, TLDR.  Well, you need to stop doing that, cut out the distractions & notifications and fake need to time-slice your attention - I cannot change the way you are, therefore I'll show you an image to grab your attention:

Oooooooo.... in grey (or gray, if you are so inclined) is GPT-4... in dark-green is Inflection-2.5.

HANG ON! that's pretty darned close accuracy, for an open-source / free LLM.  wow.

Do you want to try it out?  of course you can:

From your web-browser, navigate to: https://pi.ai/talk

I decided to have a light-hearted little chat, it is actually a really good model to interact with - if you don't like interacting with human beings on a regular basis.  I can see how young minds are going to get confused internally as they grow, not being able to distinguish between whether they are talking to a human being, who is attempting to manipulate & control them, or whether they are being influenced by a tool / application that they have mistakenly adopted to trust.  going to be interesting times....

Lovely Tony, but I want to use it on my phone... sure thing, they have a version available for Apple iOS and for Android.  As I have Android phones, I will use / setup that app, it is available from here:

As I say, personal opinions aside about the detrimental effect such tools will have on a growing human brain and the internal sociological and psychological problems that will arise from it, if you're going to embrace these type of tools, I would highly encourage check INFLECTION Pi out.

If you DID read up about Pi, you would have noticed that Inflection Pi is NOT open-source, in its own words: "I am a proprietary AI built from the ground up by Inflection AI.  It does not rely on existing Large Language Models, which allows it to have a unique set of capabilities.  Ollama (developed by Google AI) focuses on running models locally, however Pi runs on a cloud-based service that provides real-time interactions and personalised experiences."

As you can see, Inflection Pi WILL be monetised at some point - it has to be, as that's the world of business.  However, I would suggest going to check it out anyway.

I did ask the tool if its roadmap included offline usage, it stated that it has not been built for this - however, I pushed this point further and it eventually informed me that it would inform the development team that this should be included in the future development roadmap.  We'll see if that pans out.

BEWARE: this is the same as using chatGPT etc... do NOT put in any personal information or other people's personal information, the data WILL be logged and used by the internal development teams to train the models and may expose the data to other users.  Therefore, do not use it for anything that you would not stand in the middle of the street & shout out loud to everyone about - same goes for your personal information; would you really use it as a diary? no.  will you use it to record all the tasks you do in your day job no.  just THINK before you use it too much.


[final business orientated rant]: okay....so what is my problem with LLMs? well.. they were released to the General Public, ie. the consumer market - the "normal everyday people", they could access them, try them out & use them.  They, as well as businesses can now pay something like $20 per month to access chatGPT (other LLMs are available).  Those consumer market people are a LOT of people 300-800million+ users.  That's a HUGE quantity of people to test your product and to gain usage stats and training data.  You would not get that range from "Enterprise software", you'd get 1000s or 10s of 1000s.  However, by giving the consumer market access to these tools, they can use them in the workplace.  If you now attempt to "sell" or "convince" the Enterprise company to use "your LLM tools" and want to charge them millions of $$$$, there is no surprise that they reply that they can buy 5 x $20 licences for their team and they can "access AI" to help their business..... and there's the problem... Enterprise software did not arrive first, therefore it did not have time to dominate and filter down into the company.  The whole LLM release cycle has been disruptive and up-ended the normal introduction of new software.  As a business owner, why would I feel the need to pay lots of money for a Services company to help my employees to use the LLM that I've just been sold access to, when my employees are either using it themselves already or as mentioned, $20 per month access per employee, it suddenly doesn't become a good business model / market to be associated with.

I'll come back to this in a years time and see how many bubble businesses are still knocking around and see whether they have progressed into niche support LLM market areas or whether they've all gone bust - that includes the BIG players too.  Anyone remember the Yahoo! days? when a little company called Google came along and stole all their business... BIG companies do fall and I believe that LLMs are going to be the main cause of the old dead wood yesteryear companies loosing the hold and dominance in the Tech industry and maybe, just maybe we'll get some TRUE innovation back.  we'll see.

You may be wondering, why I would care / take interest in such tooling?

Well... little bit of a history lesson:

I built version 0.1 of RITA back in Apr-Jul 2018, using IBMs Watson Assistant - it was rough, it was ready, it was technology to meet a need, it changed me, it actually changed how I look at "IT projects" and how I embrace the human aspects of designing a tool that helps people.

The work I did, even made a BBC News article:

[disclaimer: my employer never acknowledged or recognised my efforts for this, even to this day - due to internal politics. sigh.] 

It looks like it has taken a while for it to go from that early version through to something in production - mainly because you CANNOT get it wrong in this subject area:

I met and discussed at length for many months with the real life Rita, the types of people she spoke to, how she approached conversations, how she stayed distanced but caring etc..etc... and I helped to build that into the first version of the tooling:

My hope is that tools like Inflection Personal Intelligence can help people, real people, like Rita did - make a difference to every day lives.... and not be driven by profit or monetisation.  

I know, I live in a fantasy world sometimes, but y'know, we have to at least try.  

Money is not really worth anything at the end of the day - but kindness & support for your fellow human beings is priceless.

Enjoy your experience with Pi.  Maybe one-day, you'll be inspired to MAKE your own "Pi" rather than having to always be a "user"....maybe...