Search Results for “headless” – Radio Free Mobile

AI Ecosystem– Google Gemini

RICHARD WINDSOR — Thu, 07 Dec 2023 06:00:19 +0000

Google gets its act together.

Google has stopped panicking (see here) and launched a new large language model (LLM) which it claims takes multimodality to a new level (true) and greatly improves the model’s ability to reason (dubious).
Most importantly, Google has productised Gemini meaning that there are different versions of it which are already being targeted at users, customers and different use cases.
This is Google’s first foray into the AI Ecosystem, and we can be very sure that at some point soon, there will be a Gemini Kit to build AI services on top of it as well as a store to distribute and sell these services.
Google has launched its latest LLM, Gemini but it has been deliberately vague regarding some of the details of what it is and its key characteristics.
Google has not said whether Gemini is a foundation model like GPT-4 or a finished product like ChatGPT or Dall-E which are versions of GPT-4 that have been further trained.
Given the benchmarking that it has done, I assume that this is a foundation model and that software development kits (SDK) will soon be available so that customers and developers can fine-tune it for their specific use cases.
Google has launched Gemini in three sizes but has not said how large each one of them is yet another sign of how the once open and collaborative world of AI has become a hotbed of fierce competition.
There are three sizes available:
- First, Gemini Ultra: which will almost certainly be cloud-based and suitable for the most complex tasks.
- Ultra is the largest model (hundreds of billions of parameters)
- Second, Gemini Pro: which Google says is the best model for scaling across “a wide range of tasks”.
- In practice, Google is admitting that to do a range of different tasks users will have to train and run multiple copies of Gemini and run them side by side.
- If it was one model that could do everything, then depending on usage, a user would only need one copy of the model to be present.
- I would guess that this model is 50bn to 100bn in size.
- Third, Gemini Nano which has been optimised for efficiency to run on-device meaning smartphones.
- There has already been plenty of noise in this space and given that the cutting edge is around 10bn running at INT4, this is roughly where I would put it.
These products are ready for the market now and Gemini Pro has already been incorporated into Bard although the Bard entity is now very cagey about whether it is using PaLm 2 or Gemini Pro to generate its responses.
However, in a very quick initial test, I have found Bard to be much more accurate than it was before even on some really obscure information as well as retrieving information from the internet accurately.
These three versions signal that Google is ready to go to market with these models and I suspect we will now see Google set out its claim to the AI ecosystem with these three at its heart.
This is the first sign for some time of what I have long believed which is that Google remains a force to be reckoned with when it comes to AI.
It made a complete hash of its message in H1 2023 when it was badly spooked by the success of ChatGPT (see here and here) and Gemini is a sign that Google has calmed down and done what it is best at.
Google has a huge installed base to which it will now roll out these products via Google Search, Gmail, Chrome, Android etc.
Gemini Nano will be made available on Pixel devices, but I suspect that discussions are already underway with Qualcomm and MediaTek to ensure that their processors are optimised to run Nano.
The net result is that Google’s new model looks like a big step forward, but it is crucial to remember that like every other model out there, it has no causal understanding of what it is doing.
It is also incapable of reasoning just like GPT-4 no matter how many times Google and OpenAI use the reasoning word in their marketing materials.
If these could properly reason, then the problem of hallucination and edge cases would be dramatically reduced and the benchmark figures provided indicate that this is not the case given how similar Gemini Ultra is to GPT-4 and how much GPT-4 can be induced to do stupid things.
Google has finally got its act together and has laid a viable claim to the AI ecosystem and if the products are as good as it says they are, then it will be in a very strong position given that 4bn users interact with Google at least once a month.
I think Google is finally justifying its position as the best AI company in the world, but the valuation of the company is pretty much reflecting this position already.
I do not have a position in Google as the valuation is not yet compelling but if there is a sharp correction or signs of life in its growth, that could change.

OpenAI & Microsoft – Power Play

RICHARD WINDSOR — Tue, 21 Nov 2023 06:50:06 +0000

Microsoft will end up with OpenAI one way or another.

The best solution to the hot mess that is OpenAI is to put everything back in its place but with the leverage that Microsoft now has over OpenAI, the acquisition is in full swing.
The soap opera continues with Microsoft now keen for the team to return to OpenAI but under certain conditions that will ensure that Microsoft has complete control of the company.
This will be marketed under the guise of AI safety and governance, but the reality is that Microsoft has no choice but to take control of this company.
Microsoft is a giant corporation and under Satya Nadella has gone from strength to strength.
Furthermore, thanks to its partnership with OpenAI and Google’s clumsy fumbling (see here), it has managed to seize the initiative in AI from Google and has been running hard with it for the last 12 months.
The result is that Microsoft will be including OpenAI technology in one form or another in all of the products from which it earns around $200bn in revenues.
OpenAI technology is the reason why Azure is now growing faster than all of its peers and is what underpins Copilot which is an LLM-powered assistant for Microsoft Office.
Hence, Microsoft has created a dependency on OpenAI technology that would have significant ramifications for revenue and profit if OpenAI were to disappear or cease its activities.
In order to fully control this risk, Microsoft will need to have full control of OpenAI, and the absurd events of the last 4 days have brought this risk into sharp focus.
This is why I have argued for most of this year (see here) that at some point Microsoft will have to acquire OpenAI and recent events have provided it with the opportunity to begin the process.
The two options currently on the table are:
- First, implosion: The board refuses to budge and almost all of the employees leave OpenAI and end up working for a new division of Microsoft headed by Sam Altman.
- This would be a disaster for the other investors in OpenAI who would be left with a pile of worthless paper and not much else.
- There would still be the issue of ownership of the algorithms and services like ChatGPT but Microsoft already has a perpetual license to all of OpenAI’s technology and so this would not affect it too much.
- However, without the employees, the odds are that OpenAI’s services would wither and die with their market share being gobbled up by a pack of fast followers.
- This is a golden opportunity for Google to wrest back the initiative from Microsoft and OpenAI.
- Second, repair: where everyone goes back to work for OpenAI, and things go back to the way that they were with a few exceptions.
- Microsoft now has the power to wipe OpenAI out as almost all the employees have threatened to go and work for Microsoft unless the board resigns.
- This means that it can make demands the biggest of which is the demand that the board resigns and the whole governance of OpenAI is recreated from scratch.
- This would most likely mean a switch to becoming a for-profit enterprise with Microsoft holding one or more board seats.
- This will be dressed up as “ensuring that these events do not occur again” and “AI safety” but it will also ensure that Microsoft’s revenue-generating products are not at risk from anything that might happen to OpenAI.
Whichever of these two possibilities ends up being the result, it is pretty clear that Microsoft will have much more control over OpenAI than it did 7 days ago.
This is precisely what it needs given the risks that it has built into its products by depending on OpenAI.
This may be enough for it to mitigate the risk to its revenues from losing access to future OpenAI technology but the best option for Microsoft is a full acquisition.
I still think that this is on the cards in the long term as this new industry develops as Microsoft will need to hang onto any real or imagined lead that it has.
Given the speed with which its rivals are already moving to lure away start-ups and services that have based themselves on OpenAI, I suspect that this issue will be dealt with very quickly.
If it drags then the ecosystem that OpenAI is trying to create (see here) will stall and may fall by the wayside.
For the last 6 months, RFM Research has considered OpenAI and Microsoft to effectively be one company when it comes to looking at players in the emerging AI ecosystem.
It looks as if this will soon be a reality rather than an opinion.

Google vs. ChatGPT – Scroogled again? pt. II

RICHARD WINDSOR — Tue, 18 Apr 2023 06:12:57 +0000

Bing chat is not ChatGPT.

Samsung is weighing up a switch to Bing on the basis that its GPT-generated search results are better than Google’s which is a proposition that I think has little basis in reality.
Google is currently the default search engine on both Samsung Android devices and Apple devices which is a reflection that in the current world of Internet search, it is the undisputed king.
This is also reflected in its market share in search which is 90%+ and it’s $224bn in advertising revenues generated in 2022.
The idea that Samsung might switch to Bing as a result of OpenAI’s superiority in GPT-based chat systems has caused another panic at Google which I take as a bad sign as Google has proved to perform very badly when it is running around like a headless chicken (see here).
I continue to think that Bing and GPT-based systems are very far from challenging Google’s dominance in search for the reasons I lay out below.
Ever since ChatGPT exploded into the public consciousness, the question that many are asking is whether these new systems make Google search obsolete.
I think that they certainly have the possibility to change the nature of the search business and that does raise some risk of a market shift, but I have not seen anything from Microsoft that would lead to believe that Bing is suddenly miles better than Google.
I have been using ChatGPT, Midjourney, Bing-powered by GPT-4 and Bard over the last few months and a few things become immediately obvious.
- First, ChatGPT vs. Bing: which are two separate and distinct services and in terms of what they do, they are very different.
- ChatGPT is a massive language model (175bn parameters) that has encapsulated the average knowledge of the internet frozen in time at the end of September 2021.
- It is very good at digging up data from its knowledge base as well as working out what the main points of a piece of writing are and highlighting those to the user.
- Unfortunately, it also has a tendency to make things up out of nowhere which means that every piece of data that it produces has to be independently verified.
- ChatGPT is frozen in time because these models are incapable of dealing with changes in the dataset and for search, this renders them useless for many of the things that people normally search for.
- The new Bing chat is a completely different animal as it makes use of the conversational capabilities of ChatGPT but when it looks for information, it does so using the Bing search engine.
- This means that the ultimate performance of Bing chat for real use cases like finding out train times or buying a new pair of shoes, it is simply the old Bing search with a conversational front end.
- Second, Google Bard: which is very similar to the new Bing chat.
- Google is using one of its smaller language models (50bn parameters I believe) to provide the conversational front-end but when it looks for information it is using the Google search engine.
- Over the last few weeks, I have used both of these systems extensively for research purposes and I find that Bard is better than Bing chat.
- ChatGPT is still much better at having a conversation or summarising text but ChatGPT is not providing a proper search function.
This is why I think that Samsung is mixing up two different products when it is thinking about what search to set as default on its devices.
If it chooses Bing to replace Google, I suspect that it will be expecting to get ChatGPT.
However, it will end up with Bing which in my testing is simply not as good as Bard when it comes to search.
Bard and Bing are pretty much the same search engines that were in use before with a conversation function stuck on the front end.
Hence, their performance will be determined by the quality of the underlying search engine and not the front end, and here Google still meaningfully outperforms Microsoft.
To make Google obsolete, OpenAI and Microsoft would need to make their GPT models able to deal with a dataset that is constantly changing which is something no one in AI has been able to reliably demonstrate to date.
Furthermore, I suspect that in order to do this, a fundamental breakthrough in the methodology of how AI is created would be needed which has not happened despite the sudden jump in terms of the quality of conversational AI.
Instead, what has happened is that a combination of Google’s 2017 transformers innovation and massive data and massive compute have been put together to produce GPT-based systems.
All of the long-standing limitations of AI still exist in these systems which is what causes them to hallucinate and make things up and this is unlikely to change anytime soon.
Hence, I do not think that Google’s search business is under any imminent threat, and should it lose Samsung to Bing, I suspect that it would come back quite quickly once it realizes that Bing chat is not ChatGPT.
Therefore, if Google’s shares suffer a large correction, this would represent a very interesting opportunity to pick them up cheap.
This has not happened yet, but it is something to watch out for.

Google – Headless chicken pt. II

RICHARD WINDSOR — Thu, 09 Mar 2023 06:25:13 +0000

Google needs to calm down and do what it is best at.

Microsoft has finally managed to get under Google’s skin and the level of panic that has ensued could see Google rush a product to market that damages the quality and performance of its core services.
Google appears to be in full panic mode with Sundar Pichai calling in long-absent founders Larry Page and Sergei Brin to help to decide what they are going to do about the new upstart technology called generative AI.
At a fundamental level, the only thing that has changed is that ChatGPT has become wildly popular amassing 100m+ users at a rate never seen before which has stoked massive media and public interest.
The underlying technology has not suddenly taken a huge step forward and in fact, OpenAI’s generative AI chatbot ChatGPT is based on a technology that Google itself invented and presented to the market in 2017.
This technology is called transformers which is a technique where the AI looks at sentences individually rather than words and calculates which other words or sentences most often occur around it and uses that as the basis for generating the output.
There is no doubt that for language, it is pretty good as it is this that allowed Google Translate to take a big step forward in its accuracy when it was updated in 2018.
OpenAI has adopted this technique and combined it with massive data sets and a lot of compute power from Microsoft to produce GPT-3 which sits underneath the ChatGPT service.
Google didn’t panic when ChatGPT was launched (as it should have done if some glaring technological shortcoming had been exposed) but only when it became massively popular implying that Google has panicked for the wrong reason.
Now Google is running around like a headless chicken working out how to include its generative AI chatbot Bard into all of its products without first considering whether it should.
The last time this happened was when Google was spooked by the rise of Facebook which culminated in the launch of the social network Google+ which was a weak product that never gained any traction.
The stakes here are much higher because if Google contaminates its core revenue-generating services with a half-baked product then its revenue, profits and share price are at risk.
Given the shortcomings of ChatGPT, I am not convinced that the panic is warranted as Google has been an AI-first company for a long time and it has successfully included this capacity into its products like Search, Gmail and Maps for years.
Consequently, I am far from convinced that Google is a sleeping giant that has been caught with its pants down but has merely failed to judge the erratic nature of what becomes popular and trendy.
I don’t think that this is Google’s Nokia moment because RFM research indicates that Google has the technology and the know-how to create better generative AI products than any of its peers but these need to be brought to market when they are ready.
The flip side of this argument is the example of Xerox which arguably laid the foundation for the computing age but failed to capitalize on it due to a lack of its management’s ability to execute on its innovation.
Google’s history in this department is pretty good although there have been plenty of mistakes, but on balance, I think Google has the ability to see this threat off but only if it does so with a calm and clear head.
Google’s rushed event in Paris demonstrated that Bard is not yet ready for prime time and gave a clear sense of the panic ensuing inside Google.
ChatGPT and Bing Chat are very far from being able to replace search yet but Bing Chat hints at how it may evolve over time.
Google is the master of understanding the long tail of search as well as knowing what the user is looking for even if he or she does not say so explicitly.
This is what has made Google the best and this advantage has not gone away whatever fancy tricks OpenAI and Microsoft have come up with.
Hence, as long as it executes rationally and calmly, I think its position as the best AI company in the world will remain secure meaning that if the share price carries on falling, an interesting opportunity could arise.
I do not have a position in Google as the valuation is not yet compelling but that may change as this hype and mayhem continue.

Google AI event – Headless chicken

RICHARD WINDSOR — Thu, 09 Feb 2023 06:29:36 +0000

Google trips over its own feet as a result of its panic.

Google is in such a blind panic that it is rushing to market with products and events that are half-baked simply to meet a competitive threat that in my opinion does not exist.
Google held an AI event in Paris combined with demonstrations online that were so full of gaffes and errors that the company’s excellence in AI was completely overshadowed.
The result was that Alphabet’s shares tanked 7.5% costing the company $97.1bn in value as the market seems to be beginning to lose confidence in Google’s leading position in AI.
By contrast, Microsoft and OpenAI have introduced pretty much the same product (although I suspect that Google’s will be way better) with worse flaws but were met with adulation as opposed to derision.
This is the problem of expectations where Google is defending its No. 1 slot and has everything to lose whereas Microsoft is the challenger with nothing to lose and of whom no one expects much.
However, this did not justify the complete mess that Google made of its Paris event which had all the hallmarks of rushing to produce something in a panic.
- First, Lost phone: One presenter was about to demonstrate a new feature when it turned out that the phone upon which the demonstration was going to be made had vanished.
- This is not a big deal in itself but it is a sign that the event was not properly rehearsed because of a sudden decision to bring the date forward from its expected date in May.
- Second, Liveview howler: During the demonstration of search within Liveview, the app incorrectly identifies a coffee shop called Copains as Starbucks (see below and see here (time stamp 25.01)).
- The Google Maps app in Chrome correctly identifies this coffee shop as Copains and there is not a Starbucks within several hundred metres.
- This leads me to conclude that either Google faked the demo (unlikely) or that the service is not yet ready and the demonstration was not adequately tested because Google was in such a rush to get it to market.
- This is a really bad mistake because it makes Google look completely incompetent with flaky products when anyone who knows anything about Google knows that it is the best AI company in business today.
- Third, Bard falsehood: A promotional video showing what Bard can do contains the factual error that the James Webb telescope captured the first picture of an exoplanet when this was actually achieved in 2004.
- Compared to the schoolboy errors made by ChatGPT experienced by myself (see here) and others (see here), the severity of this mistake pales into insignificance.
- Google’s generational AI product has all of the same limitations that Microsoft & OpenAI’s does but crucially, I suspect Google will be able to make its product far better.
- At the moment this will work in search by having the two systems completely separate in that Bard will give a response to a query at the top of the page and Google Search another that is underneath.
- However, over time, Google should be able to integrate the search graph into Bard to make it current while ChatGPT remains frozen in time in H2 2021.
- This alone will make it a far better product, but I suspect that there is more that can be done.
For some reason, the technology press has decided that Google’s product is rubbish and Microsoft’s is revolutionary even though they are pretty much the same and I suspect that Google’s will end up being much better.
This creates an opportunity because if the market loses confidence in Google’s AI, then the shares will sell off and sell off hard.
Given that almost nobody in the realm of the professional investor understands AI properly, this is a more than feasible possibility.
This is especially the case as Google appears to have lost confidence in itself and caved into the pressure brought by Microsoft to get something to market as quickly as possible.
The result was the dog’s dinner that we witnessed yesterday that felt rushed in its conception and has done far more harm than good.
I have not liked Alphabet’s shares for some time on account of slowing growth and high valuation, but now I am getting interested as the market has a habit of putting things on a big sale every now and again.
Google’s is not properly on sale yet but another 30% would make this very interesting indeed.