For those that are following via this thread. CIG had 352 Bugs & Tasks listed last week. This week they have 179. So that has meant that 173 bugs and tasks were complete in a week. Now of course it is likely that some were moved out from the table completely as things that were not really needed to hit PTU or be fixed prior to Live but still seems like it is looking good.
It has been a pain to keep informed about the info because they have changed the graphical representation of the schedule 3 times because people keep suggesting they need to show different ways. Then those same people complained when they did change it. But the principle is that all bugs, tasks and requested features are in this total list and are given a priority for getting PTU to wave 1.
Now with that, the number is likely to fluctuate up and down over next few weeks as Evocati test more things and so PTU wont be out till after the end of the sale in honesty in that I think they will be busy fixing things for the next 3 weeks to get to PTU wave 1. Now depending on what gets closed down will depend on if they get to end of December live release but that was what they were aiming for as of two weeks ago still.
Some other things to note;
Performance with multi core CPU's
The latest update to the way networking is done and some performance improvement to multicore side of things means that even clients will see performance improvement from higher core count CPU's although they haven't specified if there is a limit to core numbers then it is looking good for aiming for Threadripper styled CPU's if that is the case.
Subsumption AI
A big deal has been made about Subsumption AI but it is rather fuzzy what it really does without a lot of reading up or if you have utilised such a system yourself in development. There is a good post by LT.O''Brien on Spectrum (so thanks for that). her is the OP:
There have been quite a few posts showing confusion about what subsumption is and how it's involved in Star Citizen. Subsumption Architecture was a small part of what I've studied in computer science and robotics about 5 years back, so I figured I'd put some information down to clarify and theory build. I'd like to preface this with the fact that I am notoriously bad at fully articulating what I'm trying to explain. If there are any details missing or concepts left undefined, feel free to ask questions and I will answer as I'm able to. This is also not a detailed dive into subsumption as a concept, but more a simplified description related to SC, so there are bound to be more complex parts that I'm not going to get into.
The Idea
Subsumption Architecture is an Artificial Intelligence framework that aims to improve real-time reactions for an agent (individual problem solving entity) rather than focus on large collections of memory, based in conceptual learning. Where traditional AI tries to compare input to a huge collection of related concepts - taking so long that the state of the problem has changed, subsumption aims to see problems and solve them in real-time through simple, emergent behavior.
The Layout
Subsumption is based on the separation and completing of complex tasks by adopting (or subsuming) parts of more primitive tasks. As the lower tasks are completed, complex tasks have steps completed without specifically defining what exactly those steps have to be. This hierarchy is mostly based on the purpose of the agent and can be described as "modules" or goals. For a StarCitizen engineer this might be represented a pyramid of important goals from most complex to least:
- Work
- Fix Things
- Find Things
- Travel
- Stay Alive
The AI will take in relevant inputs to goals from top to bottom and determine, one at a time, whether that goal is relevant to the current inputs. When one is, it will begin acting on that goal until the next set of inputs. For each goal, we can add a set of relevant states, which will be used to override the goals lower in the hierarchy:
Work
-Complete Jobs
Fix Things
-Use a repair tool
Find Things
-Identify broken things here
Travel
-Pathfind to somewhere new
Stay Alive
-Avoid danger
With this list of goals and relevant actions, we can start to see how each complex goal relies on more primitive actions. Working requires maintenance jobs, to finish these jobs you need to fix things, to fix things you need to find what's broken, and so on... When an input comes in, the agent takes the shortest path through these goals to complete the simplest relevant action.
Of course this list can be brought to such a granular level that it could determine not just the major actions an agent performs, but also its personality, loyalty, and reputation. Some more pressing primitive needs may even override more complex needs in other areas. A marine may take cover when an enemy shoots, or could run away because another matter is more pressing. A hauler could deliver its cargo as expected, or choose to sell it on the black market for more profit regardless of reputation loss. It all depends on each agent's hierarchy of needs.
When our goals, inputs, and actions are defined properly, the agent should be able to quickly traverse its goals and solve them in such a novel way that it would appear wholly intelligent without a cumbersome learning algorithm or memory set that takes an unreasonable amount of time to search for a solution. The AI doesn't need to know every possible outcome, or predict the actions of others, it simply responds to inputs as they are relevant. In fact, this level of prioritization could be extended to every level of the simulation, and define even what actions the entire UEE takes as a faction.
The Implementation
Tony Zurovec and Francesco Roccucci gave a wonderful talk on what they were doing with subsumption back in July (
https://relay.sc/transcript/around-the-verse-engineering-intelligence), if you haven't watched that yet go give it a look before reading as I'll be looking at some of the specifics they bring up. And again, without more information I can only describe how I'd implement subsumption in the context of 3.0.
One thing I notice when the team mentions subsumption, they talk about systems that control much more than just the individual AI we interact with. In fact Tony says at one point that they can't push out all of the subsumption tools they've worked on because things like animation may hold it back (he even implies that the animations rely on subsumption). Since subsumption is an architectural framework for AI, it can be used on more than just the individual agent fixing a broken panel, it can be extend all the way to the 'verse wide simulation.
With 3.0 we're going to be seeing the mission givers system's first implementation, and thus the first visible use of subsumption. At this tier, the Verse's economy finds that it needs some action performed to fulfill one of its goals. The actions available to it are a plethora of job types (currently limited to combat and hauling) that can be posted for completion by both npcs and players. If we take a supply requirement as an example, we can look at how this system may be designed.
From the perspective of a single node, perhaps our Stanton 4 Mobiglass hub, we can make a few guesses at what some of these missions could look like.
http://i.imgur.com/DLZzzel.jpg
http://i.imgur.com/lR2J86c.png
We can see that Microtech requires materials from three sources (ore, liquid, and gas). These are seen coming from source points (triangles), being gathered at resource hubs (squares) and transported from there to transit hubs and manufacturers (Microtech for example) by means of the mission givers system. In this instance, our agent is Microtech, it could have various needs such as:
- Satisfy Stock Holders
- Increase Revenue | Lower Cost | Weaken Competitors
- Sell More | Increase Supply | Reduce Competing Supply
At any given time, the agent will take in information on its most pressing needs and determine the greatest level of need to be solved. With the ultimate goal of satisfying stock holders, the agent is left with multiple competing goals that can meet this need. At this point it relies on inputs to determine the best course of action. Perhaps demand isn't being met as mobiglasses aren't shipped to trade hubs, or costs are higher for ore than usual, perhaps even rival factions are simply outselling Microtech. Depending on which primitive goal is more relevant to the agent, a different mission may be generated for players to take on.
My assumption is, with 3.0, the full economic system isn't implemented, meaning there is no pressing need for various factions to actually utilize this subsumption model. There is no shifting supply and demand on each of these nodes to make one goal more relevant than another and most importantly there aren't other agents competing to complete similar goals. However, they can create the system designed to generate missions that complete arbitrary goals - at preset locations that will later be dynamic supply points - giving us a taste (and test) of primitive task completion. Later this can be expanded to include dynamic supply and competition to make tasks more varied and relevant to other activities in the verse.
Anniversary Sale
For those who are thinking about being about etc there is the sale coming up. They are going to do a lot of ATV's for the period which should cover a significant portion of information. Details so far can be seen here.
"Our Anniversary Special starts next Friday when our eight episode ship series, hosted by Chris Roberts and Sandi Gardiner, kicks off with Anvil Aerospace. Learn more about our plans, and when your favorite ship company will be profiled, by checking out the Anniversary Special
schedule."
https://robertsspaceindustries.com/comm-link/home-featured-content/16249-The-Anniversary-Special
"Hello Citizens,
We are pleased to present the schedule for the 2947 Star Citizen Anniversary Special.
Throughout the special we'll have fun activities, bring back some of your favorite ships, and include a couple of new ones. Tune in for our daily AtV videos at 12:00 PM Pacific where we'll highlight a different manufacturer each day, kicking off a new sale."
- November 24th - Anvil Aerospace
- November 25th - Aegis Dynamics
- November 26th - Ground Vehicles and Alien Ships
- November 27th - Origin Jumpworks
- November 28th - Consolidated Outland
- November 29th - Roberts Space Industries
- November 30th - MISC (Musashi Industrial & Starflight Concern)
- December 1st - Drake Interplanetary
- December 2nd-4th - All Manufacturers On Sale
So just thought I would try and give some actual useful info to those here who are interested
** NOPE! - Stelly ** 
