I almost worked with The Grid about a year ago, from my experience everyone there is nothing but kind and talented. From what I saw, the team really wants to build a great product and is extremely passionate about what they do.
Sure they have flashy marketing and throw the word AI around a lot. Clearly its working better than “Hey we use automated A/B testing, analytics, fluid auto-changing layouts, dynamic color schemes, and whole bunch of cool photo filters to tie together a unique website based on your content!”
Regarding the delays, I think the same would be true for any startup with really good pre order sales, who is feeling the mounting pressure to meet the bar they have set for themselves. This is called caring about your work.
I think the author is misguided here and took liberties on what The Grid is actually promising. I don’t understand why people feel the need to blog scathing reviews about startups they have researched for 20 minutes. Ignore the haters Grid team.
There is a line between (legal) puffery and the objective claim that they have built "AI that allows websites to design themselves."
Is it AI that allows "websites to design themselves", obviously not, even saying "a /unique/ website based on your content" (given the likelihood for small scale repetition) is a massive stretch.
Is that still just puffery or can people argue over the definition of AI? Obviously you and OP disagree but you can see his point.
The people behind TheGrid are also the same people behind Grid Stylesheets, which uses an implementation of the cassowary constraint system to facilitate web layout: https://gridstylesheets.org/
Grid Stylesheets are actually a pretty compelling approach and it was well-received here when they first shared it with HN. While the claims about machine learning and AI are arguably eyeroll-worthy, there's definitely some non-trivial engineering powering their service. They aren't peddling vaporware and it's not a scam.
I don't think OP's knee-jerk reaction here is fair—and it certainly doesn't seem to be based on an evaluation of the actual technology in question.
Questions to ask: will they eventually deliver something that roughly satisfies their claims? (Yes?) will they eventually deliver something akin to what their backers expected? (No?)
Have you backed the project, and what are your expectations? I don't agree with the author that what they promise is out of reach of today's tech, they are relatively simple optimisation problems.
I think the OP is saying he can't evaluate the actual technology in question because there is nothing to see. You are asked to take the word of thegrid folks that they have something and that they have 100 people testing it.
Yes, but OP is wrong when he says that there is nothing to see. He just didn't take the time to even look. The source code of the underlying layout engine is published on GitHub: https://github.com/gss/engine
But that's not the commercial asset that "The Grid Co" is demonstrating. They are abstracting that underlying framework, to make a more usable asset to non-technical people, akin to a SquareSpace. The people who have bought into this notion (i.e. the "founding backers") have seen almost little to no substance to establish the latter. The argument is that they are essentially being tricked. The resulting question ("is it OK to start a company this way?") is probably much more ethical in nature than anything.
As a person who is building something similar, I do know this is certainly possible with today's technology. And most probably not a scam. The team is the same team behind noflojs.org, so they do have some credibility.
But there is a misconception about Artificial Intelligence in general and its kind of true that the term is being overused. Putting couple of If/Else statement nowadays or using a Machine Learning library (Say, to find a good colour contrast - http://harthur.github.io/brain/), developers claim they are using machine intelligence and the general public might get a notion that they are doing some high-end research, while in principle they might be using one or two generally available library that any developer can use with 5-10 minutes of learning.. They really do not need to be AI experts.
By looking at thegrid.io's claimed features, I do not see anything impossible stated there. Like said by others, it's classic AI marketing by (over)using the term AI..
But the idea is brilliant though. What they are using behind the scenes are probably few algorithms to detect faces, adjust color contrast and let the user choose from a couple of fluid layout templates (formal, casual etc). The template then further adjusts itself based on the content that is provided. All they claim is automatically crop the images, find out color contrasts and adjust the typography to maximize legibility, automatically add an ecommerce widget if they see a price info etc. All of these are certainly possible and doable, even with 5-6 year old tech. Any YC level startup can do a basic, minimal version of the stated features in few months.. But the idea is huge.
Do we have any product today for a non-developer/non-technical person to build a personalized, static website without going through all the hassles? Even WYSIWYG website generators like Wix.com or wordpress/other CMS can be too much effort for a non-technical person. All they need is a randomized and personalized website where they can throw their content at and will automagically appear polished and most importantly unique, without them going through all the settings. The advantages are endless, like the themes can auto update when the industry trends change as the user never manages the design elements directly.
The author probably mistook the product as a dynamic application generator and not a static website generator, which is certainly doesn't look feasible until we have really intelligent machines..
You are right on the money when it comes to how this works. They will deliver, but it will be this, not what the blog author claims they should deliver.
"a template isn't for you. A facebook page isn't for you." ... "you could build one yourself, but designing, developing, dragging and dropping, these are all full time jobs." ... "Wouldn't it be better if websites just made themselves?" "Now they do. It's called The Grid. No templates, no coding" ... and the kicker ...
"Just tell The Grid what you want and it uses artificial intelligence to build a tailor-made home for your content."
"And tell The Grid what you want to do. Looking for more customers, more followers, higher sales, just make your selection."
The point is that the people who really understand what some silly automated website engine could do wouldn't be interested in it (they can actually make their own website as they like) and people who fund this thing don't have any idea how limited it will be, given this super slick video, which has smoothly animated layout evolution on well composed and lit three dimensional planes, and stuff like an old image of a girl playing being analyzed, points found, objects identified and then titled "Childhood", which suggests to people that The Grid will literally do that.
Of course we understand that's figurative, it won't really do any of those things... but that doesn't excuse the success of its false marketing with the wider population.
"Just tell The Grid what you want and it uses artificial intelligence to build a tailor-made home for your content."
"And tell The Grid what you want to do. Looking for more customers, more followers, higher sales, just make your selection."
I suspect that is exactly what the grid will do.
It will know a few heuristics, enough to make a site feel "personal" as in, it's not the exact template like 20000 other sites, but you can most of the time easily spot a "the grid" website.
Then again, if they can even pull that off, that's great, not at all stupid.
It was a kickstarter proj they delivered on. Don't know anything about them, and only little anout this project. However, blindly, I would def. rather back someone who has delivered a project over one who hasn't.
You should check it out and decide for yourself. I have no context to comment on the team's actual technical competence, but it has become clear to me that their M.O. is talking big and delivering fluff.
It seems to me that reacting to their marketing and reacting to their implementation approach are two different transactions. I'm a backer and am very impressed with their tech: noflo and constraint based style sheets and not only novel, they are very functional. And the idea that web publishing should abstract presentation away from content isn't new, but their implementation is very clean and emphasizes minimal friction for authoring, curating and sharing content with a self-hosted site. They have built a platform like tumblr but automation around styling and image processing. They store content as md files hosted in git and feed those into functions that do a nice job of automating presentation of that content according to some best practices and image processing that uses established AI technology. Take out the AI for color processing and you really have just a nifty publishing workflow that has built-in imagemagick support.
My understanding of their challenges in shipping have little to do with some AI grand vision and more to do with color theory. They have trained their ANN to pick the best color based on given inputs but the outputs doesn't always yield pleasing color combinations. The challenge is making colors work well with the supporting image of a content block while still being cohesive with adjacent blocks and the overall site. They should probably try solving for every style at the page level rather than trying to coordinate between blocks.
The flashy marketing video is doing its job: capturing interest and building support that leads to conversion. AI means something to somebody who is tired of wordpress and Facebook that is very different than a CS person that works on machine learning.
I hope they pull it off. I would be pretty shocked if they didn't, given the have come this far. I don't think that developers having thin social profiles is any indication that the project isn't going to be successful.
They are basically asking people to pay $96 to buy a pig in a poke. So in exchange for saving $204 (which is what a one year subscription is claimed to be offered at) when the product is "done", you are putting $96 at risk against the possibility that the product won't be right for you, or worse, that they will declare that the product has reached "1.0" status with only the barest bones feature set.
Sure, maybe the site will automagically "adapt" as you load more pictures, but beauty is in the eye of the beholder, and you might think that it looks ugly. And maybe how the site "optimizes" for particular goals is by using some canned stock phrases that you don't like and you can't customize.
Or maybe it will do the job implied by the marketing video. There's no way to know! No money-back guarantee (and if the company goes bankrupt, no recourse in any case).
Is that really worth $96? I personally don't think so. I'd much rather wait until it's done --- especially since they have VC funding --- and then decide for myself whether the resulting product is fit for purpose.
There seems to be some cognitive dissonance going on here. Almost all the posts here are saying, to paraphrase:
'It looks like the Grid team are using templates and some algorithms to to create websites while explicitly stating they are not using templates and using vague allusions to revolutionary AI for marketing purposes to attract investors and/or kickstarter funders. It is not a scam'
Misleading potential investors about what your going to deliver in order to get money is a scam plain and simple.
Come on. I don't know anything about the Grid team, but this line of criticism is pretty specious, little more than Tesler's Theorem writ large: "AI is whatever hasn't been done yet." Once AI is successfully applied to a problem, it's dismissed as not "real intelligence."
The big critique is that you were promised AI but all you're getting are "templates and some algorithms." I mean, what do you think AI is? Magic? I hate to break it to folks, but AI is just "some algorithms" applied to a problem domain.
I think there is a distinction between misleading marketing and scamming. If you promise "revolutionary AI" and only give the customer templates and algorithms, well, at least the customer have a program that can make websites. That's useful.
A scam, on the other hand, would not deliver anything at all...period. That is because, no matter how easy it is to produce the templates and algorithms, it's just cheaper to take the money and run.
Obviously, it's scummy to mislead your consumers to get them to buy a product. But I would say that misleading your consumers to buy a product that doesn't even exist is even scummier.
Send me $10 and I promise that I'll send you back $20 shortly... I might change my mind and only send you back $5 but it's definitely not a scam because I sent something back
False advertising is a scam, it's just one that lots of established companies get away with. I guess there is a difference between that and one of those confidence schemes where the entire organization is a fraud, but it's still a scam.
This seems like a good idea, and certainly do-able. Whether the people doing it can bring it off is a separate issue.
It looks like they like grid layouts. Everything is a rectangle. (On their site, a flat-shaded rectangle, like Windows 8+.) That simplifies things. You have some text and some pictures, and probably some action blocks (buttons, forms, zooms, etc.), and need a layout.
View this as an optimization problem under constraints. Use a heat map of where people usually look on screens as a basis for where the first things to be seen should go, and what needs to be on the screen at the same time. Use information about color and contrast psychology to select colors and decide how to emphasize images. They have face popout and smart cropping already, they say.
Then deliver different web site designs to different users as A/B testing, watch what happens, and feed that back into the optimization calculation. That would be fun. I wonder what would happen if you optimized for clicks. Cat videos?
Right now, there's a huge gap between the incredible complexity of CSS and what people actually do with it on most web sites. This may be a way to manage that complexity. Look at Wordpress; it doesn't do all that much, but it satisfies the needs of millions of site owners.
Thinking about it, I like the idea of using automated A/B testing to automatically drive layout optimization. That would be fascinating to watch. Provide the system with a supply of images and copy, maybe 5x as much content as the site shows, and let it optimize for maximum clicks.
The Grid's community evangelist published a video to YouTube about ten days ago that purports to show a demo of The Grid in action: https://www.youtube.com/watch?v=7vDJZ-QufBQ
Seeing so many ads for them was an immediate red flag for me. Clearly they were trying to milk more sign ups out of their promise before people started to realize it was a joke.
I don't think it is a joke. It is classic AI marketing.
From what I can tell their 'AI' chooses colors, and fonts and does some basic layout. All three bits can be done very simply. So you choose a template, it stores a few basic qualities, then uses those qualities later to choose how to parameterise its tools for adding new content to your site.
It's not strong AI, you might want to claim it isn't AI at all. But even backtracking best-first search qualifies as 80s-era symbolic AI. I don't think you need to have an undergrad CompSci to do that. My AI textbook is used by plenty of non-graduates, I certainly hope it contains enough to implement those kinds of tools without my readers needing advanced degrees or research credentials!
So I found the article overly cynical: I don't think they're claiming it will translate your human language into a site. I didn't get that claim, at least. I interpreted it as them using the inherent magicness of 'artificial intelligence' as a marketing term, confusing together things that their tool will do, with things that the AI code will do.
Overhyped product which is a big UX investment masquerading as technical IP? Definitely! A scam with no intent to deliver? I don't think so.
My first reaction seeing the video a few months ago was 'clever'. Not 'clever' as in 'clever AI or tech', but 'a neat use of very simple AI tools with a UX focus to make blogging / portfolio software a bit more flexible.'
They explicitly say that it doesn't use templates in the video. If you assume that part's bullshit then everything else is plausible, but if one of the first things they say about the produce is a lie...
Define 'template'. I'm sure there is a definition of 'template' such that they aren't using them. They're not, for example, using templates in the Wordpress sense of 'here is where blog post goes'. The video shows a selection of 'styles' which have clearly been designed, and shows the user selecting one. If you want to s/template/style in my comment, I'm fine with that.
In marketing, things are very rarely true or false. You use words in a particular sense to convey a message. You might say 'no templates', meaning 'no fixed structure templates', to convey the idea that it is very dynamic. That doesn't mean 'there is no hand-authored structure at any point', or 'the user does not chose from a range of design styles'.
Saying 'no templates' could mean 'no fixed templates' or 'we use styles not templates, a style is a general set of rules for how things are laid out, not a rigid structure'.
Not in websites, but I've build procedural content creation systems using hierarchical planners. To get really good effects they still run on what I'd call templates, they just are templates designed to work with the algorithm.
No coding, definitely, for the end user. They're not suggesting this isn't coded by someone. Just not coded afresh for each output: you build a parameterised algorithm that can generate a class of different outputs. The end user only deals with the parameters.
Anyone remember n-generate? This looks to be about the same thing a decade ahead. Honestly I kind of unironically miss that thing because with a little nudging it made okay-decent CD-covers with absolutely no effort. https://web.archive.org/web/20031119211244/http://n-generate...
Having some experience with AI, I would also be suspicious of The Grid as it is described in this article. The most inconsistent part is the natural language interpretation stated by the author, which is both very hard to achieve successfully (see Siri, Cortana and Google Now) but also never directly promised by The Grid.
There is some ambiguity of what they mean by "Throw pictures at your website and it creates a gallery" (vaguely paraphrasing) but it is a leap to say this implies language processing.
Even if the OP is overreacting and this has some substance behind it I can really relate to his disdain for the current trend of flashy landing pages with only a "subscribe to our mailing list" box, youtube videos with hipster guys extolling the virtues of a (nonexistent) product in "layman terms" and overall modus operandi of "collect emails of potential customers, raise funding and figure it out later".
What's far more likely is that they are stuck working out the kinks in the long-tail of a very complex problem, missing ship dates, and arguably not explaining this to their backers as well as they could be.
I don't use the grid, but got intrigued by their ads, and it's tech that exists now applied well. Add a photo? They'll crop it unevenly to keep the subject in the picture. Using it as a background? They'll make it less attention gathering using filters.
Yes, the term AI is overused, but they're not say, Redding University lying about the turing test. They're describing computers anticipating a need and taking measures to help meet that need, which to lay persons is very much 'artificial intelligence'.
Add a photo? They'll crop it unevenly to keep the subject in the picture. Using it as a background? They'll make it less attention gathering using filters.
Aren't these usually called "presets" rather than "artificial intelligence"?
Watching the latest video[0] on their youtube channel, it looks very similar to Squarespace or Medium. There doesn't seem to be any kind of AI or anything like that. Just an app to make blog posts with.
curious. I thought "the grid" was about using a constraint solver + A/B testing to optimize websites for arbitrary metrics - pageviews or sales or whatever. the "AI" marketing makes some sense in that context, at least.
As with most problems that can use Machine Learning (I am assuming this specific subfield of AI since they talk of the website 'adapting') the objective needs to be have a reasonable and realistic scope. The video is too vague/high-level to understand what that is.
btw. you don't need any knowledge of machine learning today to do machine learning. most clouds provide that feature. you only need to feed data to it.
Thats a sweeping statement which, in many contexts, is in error. A given ML algorithm is not effective with any kind of data or problem. Typically you need to understand the domain to pick the right algorithm and/or features and to tweak it right.
Of course, a case could be made for a product where prediction is just one not-so-significant feature, and here you can use something off-the-shelf without worrying too much about performance. But where predictions are the driver, you would need to put in some serious work to get good results. (not talking about Grid here - this is a general statement)
Currently, the absolute pinnacle of AI technology are programs which, in the highly specific contexts for which they were explicitly programmed, trick 33% of people into thinking they have the language skills of a 13 year old.
Yeah, anyone describing anything that Kevin Warrick is involved in as the pinnacle of anything* obviously wouldn't know a scam if it bit them on the face.
i am sure so sure its a scam - check out the FB page and videos of the demo etc. An ex member of the team that got fired and now causing problems ? or maybe I am the ever optimist :-)
I don't know this guy, and I haven't read any of his other stuff, so this isn't a judgement of him, it's just how I interpret the tone and content of the article itself, but...
It sounds like a rant by a corporate code-pusher who's never done anything particularly complex or sophisticated, and feels defensive about the security of his career.
Maybe that's just me. Again, I don't know this guy. I'm not passing judgement. That's just what I pick up when I read between the lines.
It sounds like there's a Kickstarter, which is targeted often to the general public, asking for money to develop a technology that eludes the pinnacle of AI research at the most well-funded organizations on the planet.
I almost worked with The Grid about a year ago, from my experience everyone there is nothing but kind and talented. From what I saw, the team really wants to build a great product and is extremely passionate about what they do.
Sure they have flashy marketing and throw the word AI around a lot. Clearly its working better than “Hey we use automated A/B testing, analytics, fluid auto-changing layouts, dynamic color schemes, and whole bunch of cool photo filters to tie together a unique website based on your content!”
Regarding the delays, I think the same would be true for any startup with really good pre order sales, who is feeling the mounting pressure to meet the bar they have set for themselves. This is called caring about your work.
I think the author is misguided here and took liberties on what The Grid is actually promising. I don’t understand why people feel the need to blog scathing reviews about startups they have researched for 20 minutes. Ignore the haters Grid team.