back to article Pimp My PowerApp: Microsoft touts AI Builder and augmented reality tools for low-code apps maker

Microsoft has crafted a thing called AI Builder, a visual tool to inject some degree of artificial intelligence into programs created using the tech giant's low-code application builder PowerApps. PowerApps is part of a set of cloud services that Microsoft calls the Power Platform. The others in the set are Power BI, which …

  1. I.Geller Bronze badge

    "Text and image analytics, previously in preview, are now generally available in Power BI."

    The images themselves cannot be analyzed, only the texts that annotate them can be - that is, found. Which was to be proved.

  2. Pascal Monett Silver badge
    Windows

    Are they that desperate to make pseudo-AI look useful ?

    It would seem that everyone and their dog has their mouths full of all this new-fangled AI that isn't AI, and the new use for this tool is counting bottles ?

    You don't need any kind of AI to count bottles. Industry has been using mechanical means since the dawn of its inception and that works fine.

    We are obviously now in Hollywood mode in this market, they're going to be redoing every simple thing that has already been done, slap the AI sticker on it and wham ! A new AI thingamabob.

    Look forward to seeing your AI coffee maker, toaster, microwave, door opener, etc.

    Humbug.

    1. I.Geller Bronze badge

      Re: Are they that desperate to make pseudo-AI look useful ?

      "...is able to identify and count stock based on a photo taken in a shop." This is AI - Microsoft AI analyses the annotating texts and finds answers.

    2. Dan 55 Silver badge

      Re: Are they that desperate to make pseudo-AI look useful ?

      There's already an LG telly with AI highlighted as a selling feature.

      I facepalmed.

      1. I.Geller Bronze badge

        Re: Are they that desperate to make pseudo-AI look useful ?

        Artificial Intelligence should intelligibly answer questions - this is the criterion. Only a human is able to intelligibly answer questions and AI, as a copy of human intelligence, must be able to intelligibly answer question.

        Microsoft AI can.

    3. Anonymous Coward
      Anonymous Coward

      Re: Are they that desperate to make pseudo-AI look useful ?

      So you're suggesting they should bring some kind of crane robot into the grocery store to count the 2 liters still on the shelf instead of just mounting a camera to do it automatically?

      1. I.Geller Bronze badge

        a new version of its Power BI

        Microsoft Corp. today debuted a new version of its Power BI. There’s a natural-language processing feature that allows users to dig into data by typing questions into a search bar. Analysts can now extend the tool’s default vocabulary with technical terms unique to their industry, company-specific terms and synonyms to help it answer queries more accurately.

        1. I.Geller Bronze badge

          Dictionary and OpenAI

          Dictionary is an universal indexer, it's structured (by alphabet) and its definitions are brief and substantive. Indexed by a few these definitions patterns can be easily found.

          What do companies like OpenAI offer? They offer their hand made dictionaries, which contain definitions only on patterns and not on words. These definitions are very indistinct and not as meaningful as that of common dictionary. (Not to mention that the use of such a dictionary can lead to 100% loss of information.)

          How to index by OpenAI dictionary outside OpenAI domain? True, it is impossible - a new database should be initiated for each particular OpenAI or another company's dictionary. With a dictionary on words, however, there is no such the problem, it's universal.

          By the way - training of data by the common dictionary on words billions times cheaper!

          Without indexing by dictionary definitions AI does not exist, and Microsoft proves it.

          1. I.Geller Bronze badge

            Re: Dictionary and OpenAI

            There is a sentence:

            --

            The judge, by the way, was the King; and as he before his crown over the wig, (look at the frontispiece if you want to see how he did it,) he did not look at all comfortable, and it was certainly not becoming.

            --

            If you use OpenAI technology you will not be able to get/ index by a dictionary definition for the idiom "by the way":

            --

            accidently (used to introduce a minor topic not connected with what was being spoken about previously).

            "by the way, pay in advance if you can"

            synonyms: accidently, by the by, in passing, en passant

            "oh, by the way Katie, you had a call"

            --

            Also you will not be able to construct these patterns:

            - the King wore crown over the wig

            - The judge wore crown over the wig

            - how the King did it

            - how the judge did it

            as well as many other patterns. And this is only one sentence! Imagine how many you lost for the whole text? And only because you decided do not see its words.

            How OpenAI parses the clause:

            -- The judge, by the way, was the King

            How many patterns OpenAI gets?

            Use my technology instead?

            1. I.Geller Bronze badge

              My database up to 99.99%

              My AI database gives almost 100% confidence that everything is taken into account and found. Training of data in other data... How good? 77%? 90%? My database up to 99.99%.

            2. I.Geller Bronze badge

              The judge, by the way, was the King

              The same sentence:

              --

              The judge, by the way, was the King; and as he before his crown over the wig, (look at the frontispiece if you want to see how he did it,) he did not look at all comfortable, and it was certainly not becoming.

              --

              AI-parsing gets out of the clause "The judge, by the way, was the King" three patterns:

              - The judge was by the way

              - the King was by the way

              - The judge was the King by the way

              3. The method of claim 1 wherein each context phrase is a combination of a noun with other parts of speech at least one of which is a verb or an adjective.

              What gets OpenAI and others, who "train data" on patterns and ignore words, parts of speech and meanings? Three patterns? One? Two?

              The parsing doesn't work without words, parts of speech and dictionary definitions! It doesn't work for the training by patterns.

              In certain embodiments, when extracting predicative phrases, the system may be configured to control for common noun phrases, idioms, or similar phrases. For example, "hot dog" may be treated as a noun, as opposed to a noun plus an adjective. Additionally, idioms such as "under the weather" may be treated as a single adjective. These noun phrases and idioms may be identified based upon a database of common phrases or idioms, but the system is not limited to any specific way of identifying them. Furthermore, in certain embodiments both the "noun phrases" and their components may be used to form predicative definitions, while in other embodiments they are mutually exclusive.

      2. This post has been deleted by its author

  3. Ben1892

    Not Hotdog

    I think I know what my "hello world" AI app is going to be

    1. Dan 55 Silver badge

POST COMMENT House rules

Not a member of The Register? Create a new account here.

  • Enter your comment

  • Add an icon

Anonymous cowards cannot choose their icon

Biting the hand that feeds IT © 1998–2019