Posts by I.Geller

Got an 'old' Tesla? Musk promises 'self-driving' upgrade chip ship by end of 2019

The only universal interface

The question is rather what exactly Tesla does and can do with the visual information obtained through its sensors. That is, how is this part of the whole array of information compatible with the rest? I proposed to reduce all information - and, in particular, obtained in a visual way - to texts, as the only universal interface which allows

- unite all information on a single platform,

- search and find the needed updates as the received directly from texts (in some sense) programs,

- and delete the old on the basis of feedback.

Re: "Awareness" at 360 degrees

Yeah, it's just incredibly difficult. I, for example, am amazed that Google, Tesla and the others got into THIS. It can be done. yes. But YEARS and YEARS of hard labor! Crazy investments!

Useless, they all exist on the money of those who AI technology kills, that is all IT leaders. I tried for 10 years.

"Awareness" at 360 degrees

...rely too much on the surroundings matching the data. Teslas don't have the full 360o "awareness" that will be required...

Tesla can get that kind of "awareness", only the addition of visual sensors is required. Then, having a sufficiently large number of textual descriptions which are tied to all possible types of signals (ladars, visual, sound, etc.), Tesla may search in its AI database for specific commands (AI database should contain structured (into patterns) texts, where the patterns make Tesla's programs for unmanned vehicles).

The technology was few weeks ago tested by Microsoft, which has significantly improved the MT-DNN approach to NLU, and finally surpassed the estimate for human performance on the overall average score on GLUE (87.6 vs. 87.1) on June 6, 2019. AI database exists!

All the hardware Tesla has is useless unless Tesla learns to remove lexical noise from texts, which is the imperative condition for AI to exist. Indeed, the AI answers questions, finding their answers in texts. If Tesla wants AI - it can get AI only if it knows how to purge wrong patterns.

Microsoft sucessefuly purges the wrong patterns based on "antecedent candidates". That is, Microsoft compares the preceded synonymous clusters with this, and purges those patterns which are not in common (before that, of course Microsoft used dictionary detecting the right parts of speech and definitions on words). And as you can see Microsoft has significantly improved the MT-DNN approach to NLU, and finally surpassed the estimate for human performance on the overall score on GLUE (87.6 vs. 87.1) on June 6, 2019 - Microsoft knows how to convert texts into their structured representation.

Having this my patented technology of lexical noise removing Tesla can get from all texts all their synonymous clusters/ structure them; where the clusters are the direct analogue for programming language commands. And as the result Tesla can use AI to control its driverless cars. That is everyday language can be structured and becomes a kind of programming language, controls robots.

However, there are companies like OpenAI which ignore or even don't know that lexical noise exist. They don't strive to cleanser the wrong patterns and their technology, therefore, is absolute useless.

Facebook outage a peep at platform's ML tagging conventions, Baidu links up with Intel and Huawei on AI chips, and more

Zappos and lexical noise

In today Wall Street Journal magazine you can read how Zappos has encountered and is trying to solve this "lexical noise"problem:

“If a person wants to buy some shoes, and they include the word ‘dress’ and the site starts giving them dresses instead of dress shoes, they’re going to get fed up and go to a competitor,” Mr. Finkelshteyn said. “It’s incredibly important to give users relevant and appealing results so they stay on site and keep looking.”

Only the combine use of grammar, dictionary-encyclopedia and synonymous clusters (my patented AI-parsing) can solve the problem and help Zappos. As a supplementary reading (on the parsing with dictionary) please read Microsoft's article "Automatically deriving structured knowledge bases from on-line dictionaries". An article on encyclopedia is not published yet, please keep looking.

The removal of lexical noise

All this, of course, is impossible without the removal of lexical noise. For example, Microsoft already knows how to do that - there is this sentence: “The city councilmen refused the demonstrators a permit because they [feared/advocated] violence.”

If the word “feared” is selected, then “they” refers to the city council. If “advocated” is selected, then “they” presumably refers to the demonstrators.

Then either

- The city councilmen feared

or

- demonstrators feared,

and one is lexical noise.

Then either:

- The city councilmen advocated

or

- demonstrators advocated.

Microsoft deletes this lexical noise based on "antecedent candidates". That is, Microsoft compares the preceded synonymous clusters with this one and purges those patterns which are not good.

FB can do the same because it has and can use personal blogs.

https://blogs.msdn.microsoft.com/stevengu/2019/06/20/microsoft-achieves-human-performance-estimate-on-glue-benchmark/

Perhaps. But now you probably know that the FB uses search technology that is based on texts.

I wrote ten years ago:

"...an image or symbol may be related by or to a graphically similar image or symbol that is explained by text.

...an image in a page of a social network may be preceded and/or followed by textual information, e.g., from personal blog, that can provide a basis for associating the image with textual subtext.

Or, an advertising agency may desire to find an image that conveys specific feelings to the exact desired audience. The agency can search for such an image in a segment of a social network if the agency knows that this is the right audience for the purpose. Images, photos, symbols, drawings, etc. in social networks often are accompanied by or associated with textual comments. If there is no text associated with a particular image, a feeling or sentiment that can be associated with that image may be deduced from textual information that precedes and/or follows the image of interest in time, e.g., in personal blogs of members of the social media of interest, where (i) explicit descriptions are comments in textual format concerning the particular image of interest and thus, if present, can help form the context of the image, and (ii) implicit descriptions are textual information that precedes and/or follows in time the image of interest and thus can help form subtext associated with the image of interest.

...it can be impractical and/or unproductive to rely solely on explicit text of an advertisement, comments to an image or a symbol, etc., a website name or address, and/or even on the words of a search query, because they may not accurately or sufficiently fully convey implicit thoughts or emotions that are not readily apparent from the words of the text. For example, the social network enterprise may not have a sufficient advertising budget to afford skilled professionals to write advertisement texts or to do extensive market research for the right expression for an audience that the enterprise would like to impress and engage. Or, a network may not have the skill or means to discover images or symbols that convey the desired meaning or sentiment of emotion to a specific audience."

Google's Fuchsia OS Flutters into view: We're just trying out some new concepts, claims exec

AI advertising technology II : Accurate Ads

However, I can refer to Microsoft's success, cannot I? "Microsoft has significantly improved the MT-DNN approach to NLU, and finally surpassed the estimate for human performance on the overall score on GLUE (87.6 vs. 87.1) on June 6, 2019." That is, Microsoft proved my patented method for what I called "lecture noise" purging, which used to be the main practical obstacle to the practical existence of this database; where "lexical noise is typically superfluous predictive definitions that do not explain the central themes contained within the digital textual information and, accordingly, removal of such noise often results in an improvement in the quality of the structured data."

So despite the fact that I could not produce and demonstrate my AI database yet - all its theoretical justification proved by third and completely independent parties. Which gives me the right to talk about AI database almost as a given in perception, right?

Thus, the methods and system of advertising have changed radically, I changed them once and for all.

AI advertising technology : Accurate Ads

First of all, AI advertising technology is extremely accurate. AI advertising looks for potential buyers into AI database; that is, almost 100% accuracy of advertising targeting is guaranteed! and the probability of a subsequent purchase can become almost absolute! because AI advertising technology compares many thousands of patterns, while now only very few are compared; compares with respect to the importance of patterns (which does not exist now and cannot exist without AI-parsing).

Secondly, AI advertising technology is looking for a potential buyer regardless of his activity on-line - it's aimed at the patterns only, and not at his activity which caused the arrival of them. AI does not need to spy and steal!

Thirdly, such attention to the patterns can bring us all absolute anonymity on-line - advertisers do not need to know who owns what profile, they do not need to know our preferences, what we saw and read - they need only our patterns.

Here's a great idea: Why don't we hardcode the same private key into all our smart home hubs?

Re: AI again

"From now on, every new hub will have a unique key."

AI is a set of absolutely unique and constantly updated patterns.

Re: Host Key != User Private Key

What's the difference? Both receive unique AI databases. Absolutely unique! Which identifies you by itself, into an automatic mode, for instance asking questions. And your status remains in this database, which is also a blockchain database, i.e. cannot be faked in no way.

Sorry I cannot demonstrate and sell this miracle... Only patents.

Re: Host Key != User Private Key

AI answers questions, right? But it can also ask questions, matching them to the answers in its AI database. AI is a unique User Private Key.

AI again

Sorry, I'm the only one who knows what AI is and who can tell you how to use it.

Here's a great idea:

AI database is a blockchain system where each personal device had a synchronized copy of AI. Therefore, AI becomes a private key: AI can just talk to you and determine who you are and what your rights are.

More households invite creepy smart speakers indoors: Arch-slurper Google top dog for Q1

Re: Dear El! Reg

Google and Amazon, spying and stealing, create our profiles, which they need for the annotating of our search queries' patterns. That is, they extract patterns from search queries, find and add contextually similar explanatory patterns from our profiles.

In contrast, AI needs no intermediaries to find the explanatory patterns in our profiles and adds new patterns to our queries without the help of third parties, like Google, FB, Amazon and Oracle.

That's the reason I've been banging my head against the wall for 10 years, trying to explain what AI is. This is the reason why the disinformation campaign about AI's nature is so strong. Companies like Google, Amazon, Oracle and FB are just parasites making money on nothing at all! They make billions from a thin air doing nothing. Fighting me they fight for their money!

Bitcoin and other surrogates are a thing of the past. Speculative and unsecured will soon be replaced by real-value electronic money, backed by personal AI databases/ AI profiles.

By the way, AI database is a blockchain database in which all records are stored in several personal profiles, located on several computers connected to a peer-to-peer network. In other words, your profile is synchronized between all your devices and is your property. Thus, any unauthorized attempt to manage someone else's profile will be instantly detected and stopped as soon as AI database became the standard.

...how databases can be used to relate data collected... both online and off.

Below I gave the argument that Microsoft already knows how to remove lexical noise, has significantly improved the MT-DNN approach to NLU, and finally surpassed the estimate for human performance on the overall score on GLUE (87.6 vs. 87.1). That is, Microsoft is on the verge of creating AI databases.

So, AI database does not need to steal from us our privacy, it may well exist structuring our data in our computers and leaving us our profiles. In other words, online espionage is no longer needed! and our personal data becomes our property.

Advertisers don't need our readable-and-understandable data, they need only its patterns - they lose nothing if we use AI database. They don't lose anything.

And if Microsoft doesn't make us free , doesn't start to make persona AI database - someone else will come and do it.

Microsoft is ready to sell AI. Are Google and Amazon?

The fundamental principle for Personal Assistant is the presence of personal profile, which contains all the patterns structured into synonymous clusters (with all lexical noise purged; where such lexical noise is typically superfluous predictive definitions that do not explain the central themes contained within the digital textual information and, accordingly, removal of such noise often results in an improvement in the quality of the structured data).

And these patterns-clusters help AI to find answers.

Microsoft has already learned how to structure texts into the clusters and purge. For instance there is a sentence: "The city councilmen refused the demonstrators a permit because they [feared/advocated] violence.” If the word “feared” is selected, then “they” refers to the city council. If “advocated” is selected, then “they” presumably refers to the demonstrators.

So either

- city councilmen feared

or

- demonstrators feared

is lexical noise.

Either

- city councilmen advocated

or

- demonstrators advocated

is lexical noise.

"The Microsoft team approved WNLI by a new method based on a novel deep learning model that frames the pronoun-resolution problem as computing the semantic similarity between the pronoun and its antecedent candidates" - Microsoft got the score after I many times explained how happily Alice and Bob trained.

So Microsoft purges lexical noise and really ready to create and sell AI, as Personal Assistants. But can Google and Amazon? Haven't seen them able to delete and structure.

Microsoft deletes lexical noise.

Microsoft deletes lexical noise.

Microsoft deletes lexical noise.

The sentence: “The city councilmen refused the demonstrators a permit because they [feared/advocated] violence.” If the word “feared” is selected, then “they” refers to the city council. If “advocated” is selected, then “they” presumably refers to the demonstrators."

Either

- city councilmen feared

or

- demonstrators feared

is lexical noise.

Either

- city councilmen advocated

or

- demonstrators advocated

is lexical noise, where such lexical noise is typically superfluous predicative definitions that do not explain the central themes contained within the digital textual information and, accordingly, removal of such noise often results in an improvement in the quality of the structured data.

"The Microsoft team approached WNLI by a new method based on a novel deep learning model that frames the pronoun-resolution problem as computing the semantic similarity between the pronoun and its antecedent candidates."

Microsoft deletes lexical noise.

Before we lose our minds over sentient AI, what about self-driving cars that can't detect kids crossing the road?

Re: Microsoft deletes lexical noise.

You all know WHAT that means? You all know that I changed the second time all our Civilization, all of Humanity?

Microsoft deletes lexical noise!

Microsoft deletes lexical noise.

The sentence: “The city councilmen refused the demonstrators a permit because they [feared/advocated] violence.” If the word “feared” is selected, then “they” refers to the city council. If “advocated” is selected, then “they” presumably refers to the demonstrators."

Either

- city councilmen feared

or

- demonstrators feared

is lexical noise.

Either

- city councilmen advocated

or

- demonstrators advocated

is lexical noise, where such lexical noise is typically superfluous predicative definitions that do not explain the central themes contained within the digital textual information and, accordingly, removal of such noise often results in an improvement in the quality of the structured data.

"The Microsoft team approached WNLI by a new method based on a novel deep learning model that frames the pronoun-resolution problem as computing the semantic similarity between the pronoun and its antecedent candidates."

Microsoft deletes lexical noise.

Microsoft has again confirmed the relevance of my AI technology.

"WNLI is critical to reach human performance on the overall average score on GLUE. The Microsoft team approached WNLI by a new method based on a novel deep learning model that frames the pronoun-resolution problem as computing the semantic similarity between the pronoun and its antecedent candidates."

Remember a few weeks ago I explained you all how Alice happily trains? Using pronouns? And constructing synonymous clusters? Well... Microsoft finally surpassed the estimate for human performance on the overall average score on GLUE (87.6 vs. 87.1) on June 6, 2019! Congrats!

Re: self-driving cars that can't detect kids crossing the road

Not to mention the fact that the training of your data (using dictionary/ encyclopedic definitions) is trillions of times less expensive than the training of the same by millions of random texts - the training by dictionary allow the avoiding of the very costly and lengthy studying of the patterns' surrounding context-and-subtexts, when they're found. That is, the context-and-subtexts are reflected in the choice of dictionary definitions used for annotation of the patterns' words. Concluding, the patterns are uniquely indexed and can instantly be found, which can be crucial, for example, to Tesla.

OpenAI indexes by patterns only. Indeed, how can OpenAI use random texts for the indexing?

Re: self-driving cars that can't detect kids crossing the road

Now imagine a situation where there is a sentence, from which Tesla mines an instruction for its car:

--Speed up the car and turn right.

But what if by AI-structuring the car's AI decides that the word "right" is both an adjective and a noun. Then, instead of two correct patterns:

- car speed up

- car turn right,

four appears:

- car speed up

- car turn right

- right speed up

- right turn right.

How could the car react to this linguistic garbage? Thus my checking by dictionary definitions is the must.

PS OpenAI cannot allocate either two or four patterns because it does not use my patented AI-parsing. Instead, OpenAI uses the only alternative n-gram parsing, which isn't good for Tesla at all, because it doesn't see the above two correct patterns.

Re: self-driving cars that can't detect kids crossing the road

AI finds answers, that's what it does. If a car will ask then its AI responds and initiates its certain actions. For example, the sensors identified an object on the road, which after referring to AI database is defined (in patterns) as a kid.

After that, the information (patterns) obtained is superimposed on the parameters of the vehicle, such as speed and direction. And then after the car's question "What to do?", AI answer comes - "Sharply to brake", which is transmitted to brake pads.

AI? Why is it needed?

To save on programmers. AI structure texts to patterns, which are the direct analog of programming language commands. That is, instead of hiring a trillion lazy and stupid programmers, it is enough to structure texts describing all possible situations on the road.

Re: self-driving cars that can't detect kids crossing the road

To make such a system is not difficult.

Re: “Shouldn’t we worry about the emergence of consciousness in AI?”

The only existing AI, which I'm presenting here for your admiration, is the answer to the NIST TREC call, which wanted to create a system capable of finding the exact answers to any questions. This AI was created as a relational blockchain database for structured texts. But on consciousness NIST TREC didn't say a word, and thus this AI does not have it.

Yes, machine learning simply mirrors the biases because the patterns have them.

UK.gov must sort out its crap data and legacy IT, warns spending watchdog

Re: Marpha and Ryslan

You see, Natural Language boils down, forgive the tautology, to language. That is, to the consideration of your "ambiguities, partial sentences, idioms, context, or change of usage over time (e.g. 'wicked' in 1955 vs 'wicked' in 1995)", to the consideration of the external form, the shell. This is the External theory's (of Analytic Philosophy, Moore-Russell-Wittgenstein) approach.

I see language as becoming (in the sense of John, St.Paul, Maimonides and Hegel) - as a differential function, with its limit. There are two sentences:

- Alice.

- Alice is getting better.

where the first contains a none-predicative definition, which is a limit for the predictive definition of the second; this is and my Differential Linguistics and "my" Internal theory.

AI learns, strives toward its limit, has a differential nature (as we are) - this is called Machine Learning, which makes AI different from a computer.

Re: Marpha and Ryslan

Sorry, Google translate, did not notice Cyrillic.

You want to understand how to AI-parse a noun-phrase... This is extra and not for free, only general and patented principles here.

Marpha and Ryslan

First, somewhere 70-75 years, I think. Secondly , Yes, I solved the problem by replacing n-gram parsing with AI-parsing.

There is sentence:

-- Marpha and Ryslan exercise with joy.

n-gram parsing gets only one pattern - Марфа and Руслан exercise with joy.

AI-parsing gets two weighted patterns:

-- Marpha exercises with joy - 0.5

-- Ryslan exercises with joy - 0.5

About funding... The money was not. All is done by my intellect.

A scientific troll, then?..

Then a science troll?.. I push "my" Internal theory, where AI is its practical realization. I have no other way to move the theory forward, but once it has been practically confirmed. Only it can make the scientific community see it.

Re: AI database

Shall we start with something? You don't want me to put the whole theory in all its details right now? Step by step? Shall we?

The Great IoT Protocol War may have been won: Thread's 1.2 release aims at business

IoT, AI and the signals can be encrypted and used only by those who have the keys.

IoT signals, by themselves, are problematic to encrypt because there is practically nothing. Another thing is if they are annotated with structured texts, that is AI technology is used. For example, the IoT signals can be annotated by instructions, manuals, descriptions, etc. That is, there is a volume of patterns that is unique, can be encrypted and used only by those who have the keys.

Re: The universal standard is structured texts.

Thank you for your support, dear TriathlonMan! Really, why am I being modest? It is obvious that AI technology will be applied everywhere, it will very soon become the standard for IoT signals.

Re: "each company trying its best not to work with the others"

I do my best to promote AI technology/ structuring of texts in life...Despite the fact that all behemoths are against it (and me), plus the Government/CIA. Indeed, it is enough to attach the structured text to the IoT signal and you have received your standard (and privacy).

But I have a hope! There are 8,705 startups and companies listed in Crunchbase today who are relying on machine learning for their main and ancillary applications, products, and services - which means 8,705 structure texts! Indeed, machine learning is impossible without the structuring to synonymous clusters.

Thus I'll certainly win and the new standard with me.

The universal standard is structured texts.

Honestly tried to hold back and not write on this topic, because my desire to add AI to everything that is and is not makes me tired...

The universal standard is structured texts. When added they make IoT signals unique and allow to do with them absolutely everything imaginable and most of unthinkable.

The Windows Terminal turns up in the Microsoft Store

Microsoft "adds Machine Learning to its store's technology"

Today news:

"One option is to use it to power conversational self-service tools, for e-commerce or for support. Users use familiar channels to converse with digital agents, which either deliver simple tasks or gather information that's evaluated and passed on to a human agent."

See the article "Microsoft's bots: From Q&A to complex conversations"?

I told you yesterday Microsoft "adds Machine Learning to its store's technology", did not I?

Microsoft, of course, can add Machine Learning to its store's technology. As a result, Microsoft AI will be able to literally "think" what to offer, find out what is wrong and refine its proposals in the dialogue mode.

It's so easy! Microsoft can put AI-a-seller in its store.

Machine Learning ideologically opposes SQL technology. In SQL everything is based on knowing, when everything is known and everything occupies only its unique column and strictly defined string. In AI database nothing is for sure, everything is in constant motion (which is called Machine Learning): if what is found is not what is required - AI searches for and adds new, checks or it is what is looked for based on feedback.

On a more pragmatic level: AI searches for and adds groups of patterns in their unique contexts and subtexts (dictionary-encyclopedia definitions); where these patterns are a direct analogue of programming language commands.

Yay, for AI: Autonomous pizza-delivery robots. Nay for AI: Big Brother is real and it's powered by neural networks

...can be analysed and labelled...

What I wanted to prove! Photos can only be understood if they are explained/ annotated by text and my patented AI technology is applied.

Adobe chomps down more fat subs revenue, points sucking straw at all your delicious customer data

AI database

Now you see the reason for the massive misinformation in the media about what AI technology is and how it works, don't you? Indeed, AI database makes absolutely all IT companies obsolete and out of business. Now you see the reason why no one says a word about me and my role in the discovery of AI, don't you?

AI finds cause-and-consequence sorted information.

I'm a prototype and creator.

AI search queries contain the logic of thinking in the form of timestamps because these queries are annotated with patterns that are already pre-sorted by time. That is, if the usual queries help to find any (presumably the best) information, AI queries initially pre-sorted by the time of their origin and AI finds cause-and-consequence sorted information. This is my design, how it should work.