Google's WebP

To compress (har har) the above message:

There are already systems out that do a better job.

Note that, like most image compression, it retains the overall meaning, but in this case it is such extreme compression that it's one sentence but contains none of the technical detail.

Uhm yes, WebM is the animation/video thingy...

And the problem this format has is more of a political and patent based problem, then a technical. It wouldn't take very much effort to implement WebM in IE/FF/BrowserX and push the update to every user. But like with WebM, many big companies are skeptical about it being patent free or they have their own solutions or they simply don't want to use some standard from someone else.
And well... 30% is of course a big difference, but given the age of jpeg and its support in the industry, 30% is simply not enough to change the industry.

But there is a reason for Jpeg and being lossy is not a big deal for most content. Png is for images with flat colors, jpg is for photos and complex images.
And you will never see a lossless format that is smaller then jpg in all cases. Even 100% jpg compresses photos much better then png and going to 95% doesn't visually decrease the image quality and you never will beat that with lossless compression.

Yeah, I've had far too many bad JPG experiences. Someone tried making the Nonja torture chamber for the map, and it didn't work because it was JPG. Back in the days of WXsand, JPG file format has broken many games.

Plus, it messes up text...

Yea I guess, but for the BIG resolutions where detail isn't important it is good. Otherwise I always stick with png...

So it looks like WebP is just another little improvement to Jpg.

I also use png for similar reasons... Oh well... lets hope that in another decade someone will come up with something better...

And I'm supposed to be an expert? I don't exactly compress images often; PNG works fine for me.

short:
Its not that hard. If the image has few colors or you absolutely need lossless compression or you have to edit it again later, then use png. If not, use jpeg at a compression level no more then 90.
If the image is extra valuable to you, store a png on your computer and upload a jpg to the web.

more:
If you use jpeg, every sane program (probably not MS paint) has a selector where you can select how much it will be compressed.
Most of the time, you have a live preview, so you can select a compression level and see if the quality degraded (too much). Usually the level is a value between 0 and 100, 90 is always a good start, most of the time, you don't need more. Many images can get away with 80 without loosing too much detail.
Here is Memzaks avatar with jpeg-75:
disabled.zzl.org/memzak-75.jpg
disabled.zzl.org/memzak.png (original, for better comparability, I added a background color.)
The jpeg is 10,504 bytes, the png is 66,697 bytes. So you have a compression ratio of over 6 compared to png (the uncompressed image would be over half a MB)
And yes, if you zoom in and look very closely, you will see differences. But thats not the point. I'm not sure how you surf the web, but I don't do it at 4x zoom. If you can barely see it, everyone else who doesn't look closely will not notice the difference. And thats what jpeg is about.

extra:
Oh and another thing about compression: If you use images on the web, scale them appropriately. Memzaks avatar is viewed at 100x100 pixels, yet it is stored at 451x401. If stored at the proper resolution the images are:
disabled.zzl.org/memzak2.png (8,360 bytes)
disabled.zzl.org/memzak2.jpg (2,212 bytes)

Advanced info:
Like I said earlier, there is something called color subsampling. Every video (you know of) is stored in a way, where the color information is only stored at a quarter the resolution as the luminance information. (So they are not in the RGB colorspace.) Jpeg knows a similar scheme. Jpegs are not stored in RGB colorspace either, they also have one luminance and two chrominance planes. The color planes can be stored at different resolutions. For most natural images, its enough to store the color at quarter resolution. Higher color resolution is only needed, if you have fine sharp transitions between different colors or fine details in colored regions. This is more visible in small images, because if you look, you usually look closer.
The first jpg above was stored with quarter color res, the small 100x100 version is stored with full color information. As a comparison, here is a quarter res version:
disabled.zzl.org/memzak3.jpg (2,230 bytes)
If you look closely, you will see while it still looks very similar, there is a noticeable loss of detail.
You don't have control of the subsampling in every program. I use Gimp for my image editing and there you have full control about the subsampling in the advanced options. I don't remember seeing that option in Paint.Net and I don't remember it in photoshop either, but I didn't use photoshop for (10?) years. Still photoshop should have it.

*edit*
extra for png:
Pngs are compressed with the same compression algorithm zip uses (deflate). So you might know in many zip programs you can set a compression level too. Better compression levels just use more time to compress. While only few image editing software allows to set a png compression level, there are standalone programs you can use. Gimp again compresses pngs pretty good. But sometimes you can get as much as a 50% size reduction losslessly.
I can't really recommend a software here, I use pngout, but its a command line program, so you probably don't want to use it. For more information, go here

Ooh, that IS interesting. I guess I can use JPG on some things.

Of course, it should only affect loading time the first time you view it since it's cached after that, but still, good for memory savings.