#AI text recognition

[johncarter54]

Write essays and do your homeworks much faster by using AI powered tool NetusAI. Less work - more time for your life!

[globosetechnologysolutions1]

The journey of building an OCR training dataset—from data collection to model training—is essential for creating reliable and efficient text recognition systems. With accurate annotations and stringent quality control, businesses can unlock the full potential of OCR technology, driving innovation and productivity across industries.

[manasseh]

auto translation my beloved. center of the fava bean it is

[strangeite]

[ID: First image is a headline that says, "Survey Says: 'Baltimorese' is among the hardest accents in the nation for AI to understand"

Second image is a screenshot of tags by runawaymarbles that say, "#what's this?? it's Aaron with an iron urn!!" /end ID]

Baltimore is a beacon of hope in the war against The Machines

[engenhariadesoftware]

As Melhores IAs de Conversação com Fala Gratuitas

Introdução às IAs de Conversação com Fala Nos últimos anos, as IAs de conversação com fala têm ganhado destaque em diversas áreas, desde assistentes pessoais até chatbots empresariais, passando por sistemas de automação doméstica. Esses sistemas utilizam tecnologias avançadas de reconhecimento de fala, processamento de linguagem natural (NLP) e síntese de fala (Text-to-Speech) para permitir uma…

[neturbizenterprises]

Revolutionize Tech with Multimodal AI!

Multimodal AI is revolutionizing technology by seamlessly combining text, images, and audio to create comprehensive and accurate systems.

This cutting-edge innovation enables AI models to process multiple forms of data simultaneously, paving the way for advanced applications like image recognition through natural language prompts. Imagine an app that can identify the contents of an uploaded image by analyzing both visual data and its accompanying text description.

This integration means more precise and versatile AI capabilities, transforming how we interact with digital content in our daily lives.

Does Leonardo AI, Synthesia AI, or Krater AI, leverage any of these mentioned Multimodal AI's?

Leonardo AI - Multimodal AI:

Leonardo AI is a generative AI tool primarily focused on creating high-quality images, often used in the gaming and creative industries. While it is highly advanced in image generation, it doesn't explicitly leverage a full multimodal AI approach (combining text, images, audio, and video) as seen in platforms like GPT-4 or DALL-E 3. However, it might utilize some text-to-image capabilities, aligning with aspects of multimodal AI.

Synthesia AI - Multimodal AI:

Synthesia AI is a prominent example of a platform that leverages multimodal AI. It allows users to create synthetic videos by combining text and audio with AI-generated avatars. The platform generates videos where the avatar speaks the provided script, demonstrating its multimodal nature by integrating text, speech, and video.

Krater AI - Multimodal AI:

Krater AI focuses on generating art and images, similar to Leonardo AI. While it excels in image generation, it doesn't fully incorporate multimodal AI across different types of media like text, audio, and video. It is more aligned with specialized image generation rather than a broad multimodal approach.

In summary, Synthesia AI is the most prominent of the three in leveraging multimodal AI, as it integrates text, audio, and video. Leonardo AI and Krater AI focus primarily on visual content creation, without the broader multimodal integration.

Visit us at our website: INNOVA7IONS

Video Automatically Created by: Faceless.Video

[tepperz]

It shouldn’t be called artificial intelligence – AI. It should be called artificial patterning – AP. There is nothing intelligent about it. It’s only pattern recognition, and basic, flawed recognition at that. It’s not Intelligent art, it’s Patterned Art. It’s not Intelligent writing, it’s Patterned writing. I think a lot of the optics around the “AI” craze can start to be broken down with a transition to calling it “AP”.

[ehmaree]

That's correct. Recaptcha was actually designed to have two purposes. The first purpose was to make sure bots weren't visiting those sites to fill out the information (obviously). The second purpose was to improve the text reading ability of the algorithms in use. And the way the original Recaptcha did this was quite interesting. Books and texts would be scanned and uploaded en masse for archiving purposes (and to archive something is a beautiful purpose indeed), and when the algorithm couldn't find what a word was, it would be sent out in these Recaptchas for people to find out what it was.

Now, Recaptcha has evolved quite a bit over the years, but its purpose is still more or less the same. Root out the bots while training the algorithms.

[adasitecompliance]

Digital Content Accessibility

Discover ADA Site Compliance's solutions for digital content accessibility, ensuring inclusivity online!

[globosetechnologysolutions1]

The journey of building an OCR training dataset—from data collection to model training—is essential for creating reliable and efficient text recognition systems. With accurate annotations and stringent quality control, businesses can unlock the full potential of OCR technology, driving innovation and productivity across industries.

[makingqueerhistory]

":')))))))) you realise that gen AI is available to everyone though right??? Queer creators can use it just as much as anyone else??? I just don't understand this post... It really feels like a cheap way to get on the 'AI Bad's bandwagon, and coming from such a thoughtful and insightful creator that's incredibly disappointing... It's okay to not comment on subjects you're not an expert in y'know...?"

Y'all know the drill, I am replying to this publicly but that is not an invitation to send any negative messages to the person I am replying to.

Anyways, let me start by saying that the original context of the post you're replying to is discussing an event where a queer org used generative AI to steal an interview with Keri Hulme. So let's start there. To be clear I don't even know if the original interviewer was queer so let's put the identities of stealer and stolen from to the side. I want to explain the harm done in this example specifically and I hope this is illustrative of what harm generative AI can (and does) do.

The original place I saw generative AI was a queer org that explicitly says they are using generative AI "for good", and as a way to bring more queer history to light. So let's take them at their word, and assume they are not out to cause harm. This is the best example of generative AI that I can imagine, so I hope that makes it clear that I am not coming at this issue from bad faith in any way.

Here is the harm they are causing:

Decontextualizing and rephrasing an interview: I am not going to pretend that I am an expert in academic best practices, but I do believe one thing, if a person is speaking on their own identity and lived experience, it is always much better to directly quote than it is to rephrase. As I read this source, I initially didn't know that it was AI, and I was already upset. An interview that is widely available on the internet with no pay wall, was poorly sourced and made more vague than it was in the initial text. By creating one degree of seperation between the original words of A WRITER (whose literal job was largely based in choosing the right words to describe experiences they had) harm is already done. It makes vague what was once clear, and removes Keri Hulme's voice from her own narrative.

The original interviewer is not paid, or given proper recognition: I get it, sometimes just copy pasting an interview doesn't feel transformative enough, but something that one would learn if they worked in the queer history field and weren't a literal robot rehashing what has already been said, is that not everything needs to be transformed. In those cases, we give credit to the person who said the original words (in this case Keri Hulme), and the interviewer who facillitated the conversation (in this case Shelley Bridgeman). This case (again a best case scenario), takes the attention and byline away from the original interviewer and gives it to an AI.

The original publisher of this story is deinsentivised from paying interviewers in the future: The original publisher of this interview has ads on their website. As a person who also has ads on their website, taking an article like this and rephrasing it for no good reason (the orginal word count was not prohibitive and the rephrasing did not make it more readable), takes money from the publisher. It's pennies, but it's also removing numbers could have been used to justify further interviews with asexual people and archiving of asexual stories. The org that stole from this publication does not interview people themselves so the money and numbers that could have gone to continue to preserve asexual stories goes to stealing them instead.

These are just the active harms that I saw in this specific case. As you said, I am not an expert in generative AI, and will not be speaking as if I am. But I will say that asking me not to speak out on active harm that is being caused in queer history spaces, is disrespectful to my many years in this field.

To illustrate this even clearer: if you were a patron, you would know I recently took down an old article. I have been rereading and editing our backlist of articles, and I found one that no longer fit my standards of sourcing. My standards had recently raised due to a video made by HBomberguy about someone in the queer history space who was stealing from other creators. I watched this video not as a work project, but because I watch most of HBomberguys videos, and this one made me think more critically about sourcing. An AI can't do that. All an AI has is what has been inputted, and it is right now impossible to input every available peice of information about ethics into an AI and get a coherent ethical basis on which it will function.

It is a distinctly human trait to absorb information and change in that way. AI can rephrase information that already exists, steal it, recontextualize it even, but it cannot create something altogether new.

Do I believe that there one day might be an ethical use for Generative AI? Maybe. Do I believe that coming into a queer history space, stealing the words of a Maori asexual author, rephrasing them, and giving the original interviewer and publication no form of compensation for their work, is accomplishing that? No.

On a more personal note: I am coming at this issue with a bias. As a queer history creator, I do not want AI in my space, because it is literally damaging to my financial prospects. It has been like pulling teeth to try and get patrons in the current state of the global economy. I don't blame anyone from that, but I feel very disrespected that I am being asked to compete with a machine now. Not only that, but I am being asked to shut up and be fine with it? No, absolutely not. I cannot and will not stay quiet as space that I have fought tooth and nail to create in mainstream discussions is taken and given to AI.

AI was not supporting me when I was sent gore to try and scare me off of discussing queer history. A person did that. AI was not there to tell me I had written too many sad stories, and I needed some happy endings to remind myself of the good in the world. A person did that. AI was not there when I was being harrassed for supporting and including asexual stories on my website. A person did that.

And after all that, I am being asked to lie down and take it when my ability to pay the people who supported me in those ways, is being threatened. Nope. Not going to happen.

An AI doesn't have to make rent. An AI doesn't understand what it feels like to have to stop holding their wife's hand in public. An AI didn't get calls from people needing comfort in reaction to the election. Pay me for my work, and get this AI nonsense out of my face.

[prokopetz]

Okay, so you know how search engine results on most popular topics have become useless because the top results are cluttered with page after page of machine-generated gibberish designed to trick people into clicking in so it can harvest their ad views?

And you know how the data sets that are used to train these gibberish-generating AIs are themselves typically machine-generated, via web scrapers using keyword recognition to sort text lifted from wiki articles and blog posts into topical subsets?

Well, today I discovered – quite by accident – that the training-data-gathering robots apparently cannot tell the difference between wiki articles about pop-psych personality typologies (e.g., Myers-Briggs type indicators, etc.) and wiki articles about Homestuck classpects.

The upshot is that when a bot that's been trained on the resulting data sets is instructed to write fake mental health resource articles, sometimes it will start telling you about Homestuck.

[indian-school-of-ethical-hacking]

Smart AI Chatbots understands customers' language and expressions and has real-time solving power with accurate answers. Explore its benefits and limitations.

[industrydesignservices]

Natural Language Processing (NLP) is a part of artificial intelligence that allows computers to recognize natural language – the words and sentences that humans use to communicate – to generate value. While machines are excellent at operating with and understanding structured data (such as spreadsheets and database tables), they’re not so great at deciphering unstructured data, for instance, raw text in English, Polish, Chinese, or any other human language. To know more about browse: https://teksun.com/ Contact us ID: marketing@teksun.com

[askablindperson]

In what way does alt text serve as an accessibility tool for blind people? Do you use text to speech? I'm having trouble imagining that. I suppose I'm in general not understanding how a blind person might use Tumblr, but I'm particularly interested in the function of alt text.

In short, yes. We use text to speech (among other access technology like braille displays) very frequently to navigate online spaces. Text to speech software specifically designed for blind people are called screen readers, and when use on computers, they enable us to navigate the entire interface using the keyboard instead of the mouse And hear everything on screen, as long as those things are accessible. The same applies for touchscreens on smart phones and tablets, just instead of using keyboard commands, it alters the way touch affect the screen so we hear what we touch before anything actually gets activated. That part is hard to explain via text, but you should be able to find many videos online of blind people demonstrating how they use their phones.

As you may be able to guess, images are not exactly going to be accessible for text to speech software. Blindness screen readers are getting better and better at incorporating OCR (optical character recognition) software to help pick up text in images, and rudimentary AI driven Image descriptions, but they are still nowhere near enough for us to get an accurate understanding of what is in an image the majority of the time without a human made description.

Now I’m not exactly a programmer so the terminology I use might get kind of wonky here, but when you use the alt text feature, the text you write as an image description effectively gets sort of embedded onto the image itself. That way, when a screen reader lands on that image, Instead of having to employ artificial intelligences to make mediocre guesses, it will read out exactly the text you wrote in the alt text section.

Not only that, but the majority of blind people are not completely blind, and usually still have at least some amount of residual vision. So there are many blind people who may not have access to a screen reader, but who may struggle to visually interpret what is in an image without being able to click the alt text button and read a description. Plus, it benefits folks with visual processing disorders as well, where their visual acuity might be fine, but their brain’s ability to interpret what they are seeing is not. Being able to click the alt text icon in the corner of an image and read a text description Can help that person better interpret what they are seeing in the image, too.

Granted, in most cases, typing out an image description in the body of the post instead of in the alt text section often works just as well, so that is also an option. But there are many other posts in my image descriptions tag that go over the pros and cons of that, so I won’t digress into it here.

Utilizing alt text or any kind of image description on all of your social media posts that contain images is single-handedly one of the simplest and most effective things you can do to directly help blind people, even if you don’t know any blind people, and even if you think no blind people would be following you. There are more of us than you might think, and we have just as many varied interests and hobbies and beliefs as everyone else, so where there are people, there will also be blind people. We don’t only hang out in spaces to talk exclusively about blindness, we also hang out in fashion Facebook groups and tech subreddits and political Twitter hashtags and gaming related discord servers and on and on and on. Even if you don’t think a blind person would follow you, You can’t know that for sure, and adding image descriptions is one of the most effective ways to accommodate us even if you don’t know we’re there.

I hope this helps give you a clearer understanding of just how important alt text and image descriptions as a whole are for blind accessibility, and how we make use of those tools when they are available.