-
Posts
430 -
Joined
-
Last visited
Content Type
Profiles
Forums
Developer Articles
KSP2 Release Notes
Everything posted by snkiz
-
I Have never seen so much copium over such an abstract subject. Impressive how you completely ignored my answering my own question on the fate of Watson. Seriously Do some reading. The people creating these things are crazy bok smart, but clearly don't have children, have never had to idiot proof a system. And since ai algorithms are black box. They don't know how to fix it.
-
Ah, further reading. Watson's failure was a product of garbage-in garbage out. They, took a shortcut with synthetic data sets from a singular source of doctors at one hospital. It also suffered bad ux design that led to mistakes in entering patient data, and was apparently unwilling to say "I don't know, this requires further review." It had failure rates upwards of 50% What that says to me, is in a decade we have learned nothing about ai. And worse these really smart people are making basic mistakes that any practical person would not have made for their given field, or just common sense in some cases. So that was beyond overselling, Watson was bordering on a grift.
-
That's right I remember reading that was it's new job, and it finding things doctors had missed. I was not aware it was pulled. As I remember it did this by comparing an inhuman number of cases, cross referencing symptoms and such. At the time that seemed like a good idea to me, giving doctors this database they couldn't possibly keep in their head. Did they let the thing run unsupervised? or were doctors over trusting in it? Medical is messy, with lots of overlapping things for conditions that are not related at all. In those conditions can it really be expected to give definitive answers? Or operate more as another diagnostic tool, to just look at possibilities? I can't imagine it would have failed that severely in that role. Or maybe IBM just oversold it.
-
I actually feel like this is a step backward. It was only a few (maybe a little more than a few.) years ago that IBM cleaned up on Jeopardy. Whatever happened to Watson? Or more importantly it's algorithm? They like to point out it's a conversation ai, and not a math ai. No one wants to chat with a machine that going to constantly blow smoke up your know where. What is the point of a conversation ai designed like this, without any fact checking ability? In some cases not even basic math skills?
-
Thanks for the reading, and the follow up. My side still hurts from going down the peanut butter sandwich rabbit hole. It's hilarious that there is already a word for hacking them. "prompt injection" I'm beginning to think there is no shortcut to training an AI. They don't even have the intuitiveness of a child. From reading that it would seem the only way AI has a chance of working right now is if you, assume it's dumber than a bag of hammers. Painstakingly teach it as you would a child, and constantly check its work. Give it grades. They did hit one nail on the head, AI has no BS filter, and no one seems sure about how to implement one.
-
I hadn't thought to correct it.
-
I know it's round. But the decoupler hardware gives the inside silhouette flat edges, like the mk3. At least that was the first thing that came to my mind when I saw it.
-
It did not go well. First I asked it how to play kerbal space program. It gave a 6 step, kind of general overview of what kerbal is about. Named dropped some planets, OK so far. Then I asked it How much delta-v to get to minmus? It gave me the incorrect delta-v for the transfer burn, (500-600 m/s) It also didn't mention the inclination at all. It did say it depends on where you start. Mentions in game tools (like a detla-v map?) and Mechjeb. It also incorrectly asserts that the Dv required is dependant on your mass and thrust level. I may need to be more specific. Ok, How much Dv do I need from the launch pad? This is where it gets interesting. First it again repeats, it depends. This time on altitude and velocity, I think it expects me to air launch, 600-700 m/s this time. Then it goes into further detail. It tells me to launch from the launch pad to LKO, (it did not use acronyms.) at an altitude of 200 km for about 7,500 m/s! Now it mentions the orbit adjustment burn, but doesn't call it an inclination change. "A significant amount of 200-300 m/s". Next comes the transer burn, point for using the words Hohmann tranfer. However this burn is also 200-300 m/s. And finally it correctly labeled the 'Orbital insertion' burn But gave me a value of 50-100 m/s. Here it says again, it depends, but you can calculate what you need from the physics engine or use "In game tools" such as DV map or mechjeb. So this isn't going well. One more, I ask it "Can you guide me to to build a rocket capable of landing on minmus, with the Dv required?" It goes on to give me some really generic "tips" on what parts to use. It does say to consider things like Dv required, mass, and stability. It says a typical landing minmus mission is 800-1000 m/s. this time it combines launch, inclination change, and transfer into one step. and tells me I need 600-700 m/s to reach minmus from the launchpad? it goes on to say now I need to perform my landing burn and enter Minmus's atmosphere? Doesn't say at all how much dv. And only says to save enough to get back to Kerbin, without saying how much that is. All and all it was very convincing, and wrong. Oh so wrong. Thing is, if I didn't already know what I was doing, I might have tried to follow it. I mean with 7500 dv I'm sure I would have figured it out. But it was erie how confidant it was, the details it did manage to include. I personally agree with ChatGPT. A Dv map and Mechjeb (or something like it.) should be in game tools. This was a fun little distraction. It has me much less worried about it as a tool for plagiarism, research, or even a search tool replacement. If Microsoft integrates this into bing, it's just going to be an even bigger joke than it is now.
-
I get the feeling this isn't a serious issue to you.
-
Is it just me or does the booster look like a MK3 tank from above?
-
I saw one the discord mod explain the reason for no voice chat is the the ESRB rating. So the game probably won't have voice chat. Witch is for the best.
-
Joystick wish
snkiz replied to ApolloEng's topic in Prelaunch KSP2 Suggestions & Development Discussion
You and a bunch of people, welcome to the forum. -
KSP2 Insiders - Preview event hosted by ESA
snkiz replied to Vl3d's topic in Prelaunch KSP2 Discussion
my bad Marketing BS, High profile event, of course they rolled out the top sku. And the RGB makes it go faster. -
Asynchronous blinking would imply they are, or drunk.
-
KSP2 Insiders - Preview event hosted by ESA
snkiz replied to Vl3d's topic in Prelaunch KSP2 Discussion
Some german PC builder sponsored the event. -
Kerbal Space Program 2 Release into Early Access Feb 24th
snkiz replied to Intercept Games's topic in 2022
Purely an optimization problem. Even the new consoles are fairly weak compared to a PC. But the consoles have quality standards, dev machines that give them inside access to the hardware, and they don't have to deal with windows. Games that are console first are often not optimized at all for PC, minimum effort. PC first games must have more attention to fit the very specific hardware, and it still doesn't always go right. A good multiplatform game is to generic to fully use the any hardware, unless many builds are done and lots of money is spent. -
hence the word IF. it's all speculation at the moment.
-
There's a button to turn the windows transparent. The issue then was what that does is render the IVA through cutouts in the model. This is why it made sense to make that optional. That's a whole other model and textures to load, that you aren't even looking at a lot of the time. The new windows look great, but if I need those 10fps that costs then I'd like the option to turn it off.
-
I assume the existing one will be updated https://wiki.kerbalspaceprogram.com/wiki/Main_Page
-
I'd love to jump in day 1 but I have to consider priorities. I may have to wait a few weeks.
-
I swear the next screenshot is just going to be a fog bank to quiet this thread.
-
something new to ponder https://twitter.com/KerbalSpaceP/status/1623728427470671872?s=20
-
Concern about procedural things and other 'QOL' improvements
snkiz replied to snkiz's topic in Prelaunch KSP2 Discussion
No it doesn't appear that way. Although I really don't think this would have been the success that it was if it didn't start out silly. That might just be it, our little kerbals are growing up.