The Grid: Certainly Misguided, Probably a Scam
benhorrix.com
benhorrix.com
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.
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.
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.
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.
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..
"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.
"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.
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.
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.
If that's the case - why don't we have any open source content management systems that will do what The Grid does already?
'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.
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.
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.
Scam?
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.
Someone could probably get VC funding for that.
And here's the site whose content he's populating in the video: http://myothercamera.is
My takeaway from this is that The Grid has a fantastic marketing team, and a product that may be a decent competitor to Squarespace. Maybe.
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.'
[edit: add last para]
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...
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'.
So no, I don't assume 'no templates' is bullshit.
"It's called The Grid. No templates, no coding."
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.
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.
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'.
Aren't these usually called "presets" rather than "artificial intelligence"?
[0] https://www.youtube.com/watch?v=7vDJZ-QufBQ
Edit: and here is the final result http://myothercamera.is/
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)
No.
*Other than possibly hype generation
Clearly the venture capitalists would have uncovered if the product was a scam during due diligence.
Right? :P
Yes.
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.
Also, it's called perception. My comments were in reference to the tone of the article, not the author.