Chapter 43

After the sharing of the preliminary vision for the future, life on the ship developed into a routine. The crew decided that their space ship would hence forth be referred to as ‘Dóchas’. The naming of their future Mars settlement was put off for the time being.

One evening as they finished their usual discussion time, several of the crew expressed frustration with the learning stations. In probing to learn the source of the difficulty, Sarah’s Clan found that some crew members were having problems interacting with the learning stations telepathically. Several noted that for them it was like moving from first learning a new language, to being comfortable with it. With this in mind, those who were comfortable with the telepathic aspect of the learning stations were designated to be coaches. The effort immediately began to show multiple benefits for everyone involved. Those gaining the improved skills were more productive and confident, which showed in their attitude. The coaches discovered that their efforts strengthened their abilities, which even carried over to their family’s mind-link.

As the crew’s knowledge level grew, their focus on learning shifted from simply repairing or refurbishing an item to consider its effectiveness. This evolved into proposing modifications to the group. Nearly always after a careful review the item was modified so that it filled its role more effectively.

It was during this time that they altered their work schedule to have all sections staffed around the clock. To do this they elected to have ten hour shifts, which gave them an hour overlap at shift change. This gave them the opportunity to discuss progress and go over pending issues. Another benefit was that it allowed them to do tasks requiring more people. One downside was that not everyone was physically able to be at the after breakfast or after dinner discussion sessions. The solution was to attend electronically and telepathically when on duty. By using their telepathy, their participation was nearly as effective as being there in person. The downside was they missed the physical aspect of being together.

It was about a month after the presentation of a vision for the future, that the Computer Engineering Department scheduled a discussion on updating the AIs.

When the Command Staff entered the department Gillian said, “Good morning.” All of her coworkers chimed in.

“Good morning to all of you,” replied the Command Staff. They quickly greeted everyone before sitting down.

Gillian looked around nervously then said, “The first part of our meeting is to go over the issues the AIs are currently encountering.” Seeing the Command Staff’s expressions she added, “These issues actually impact all the AIs. Star’s is the most obvious and serious situation due to her role. We think, we’ve found a solution that will benefit us long term. Marléne will take the lead on covering those issues.

“We have two proposals. One is a conservative approach and is basically a refurbishment with some significant changes in software. The second is a complete replacement of the hardware with new programming. This has the potential for a much greater gain but it has some significant risks. Wang will outline the proposal that modifies both the hardware and software. The changes are focused on resolving current issues. Li Min will give the presentation for the replacement proposal. We believe it is the best long term solution, presuming no unrecoverable gotchas.”

“Thank you,” said Marléne. “We’ve analyzed the AIs current processing architecture, logic flow and memory structure. This includes the programming issues they brought to your attention some time ago that were addressed. We are impressed by their ability to define the cause of their issues and their proposed solutions. The changes and implementation resulting from that effort were well done. Best of all they improved their performance.” Seeing several about to comment she smiled and added, “When we came to that conclusion, we didn’t know that several of you were involved in the improvements.

“Even with those changes there are still significant constraints in both the hardware and programming. These are primarily due to design limitations, but there are a growing number problems due to aging.” She then went over the key issues and noted where they affected multiple AIs. “In summary, we don’t think any of these issues are critical when viewed individually, however, if they cascaded we could have a difficult time. We could cope but there are gaps that the bots fill better than us bio’s could, but we really don’t want to be in that position. We believe we should do everything possible to avoid the circumstances that would put us in that position.”

At a nod from Marléne, Wang said, “There are many improvements we could make in the processing instruction set. By doing this, we could correct some design issues and improve the instruction set to yield an increase in performance and response time. Making any changes to the processors is not feasible. There are however, bottle necks in moving data to and from them that this proposal would focus on reducing. We estimate all of this will gain us about a factor of two or three in improved performance. To accomplish these changes, we will essentially violate some of the system’s basic logic rules, which could have some significant risks associated with doing so. One concern is that they may not always function as planned, especially when they are operating near maximum utilization.”

“I am not surprised,” said Rusty. “When we looked at the solutions we implemented earlier, we saw routines that had us wondering ‘why was it done like that’. However, that is often the case when you go back to look at routines you’ve written.”

“So where do you want to go with this?” asked Sarah.

“Ah, thank you,” said Li Min. “In the archives, we found a design for a multilevel, multi-core processor. It is configured as a 3D array. Obviously it is very complex but the documented benefits were outstanding. The design and manufacturing information is very detailed. When we found it we were surprised that it wasn’t finished. In digging through the project documents, we found that there were issues during the late stages of development. These occurred, when it was in advanced prototype evaluation.

“This was in the same time frame that the current processor – memory structure design was developed and brought into production. This is the current design in use and it follows previous designs, while the 3D array appears to be an out of the box type of change. So as you’ve probably guessed, our curiosity was aroused. We did find some test data on the 3D and individual layers functioned well, but never when it was fully assembled. The developers tried several solutions, and some multiple times, none were successful. The end result was that the development project was essentially shelved. As far as we can tell, they never figured out why.

“After reviewing the data, we think this was due to the layers not being correctly aligned. Considering the interface’s complexity, it is something that could easily occurred due to the tools they were using. In the last two days, we’ve built two of these processors in a virtual environment and they work fine. So we would like to build a couple of real ones.”

“So how are you going to solve the alignment issue?” asked Sally.

“With index pins.”

“You mean there weren’t any.”

“There were, but we think they weren’t unique enough. When it was assembled, it appears that two layers could be placed out of sequence and another could be if it was rotated around the z-axis.”

“It might be nice to know how they missed something like that.”

“We agree.”

“And what other changes are you making?” asked Claudette.

Several laughed then Gillian said, “If we go with the 3D array, just about everything. However, it is our conclusion that the only thing keeping it from working is putting the layers together correctly. We have all the detailed design information for the interfaces to support this processor, along with just about everything else. To us it is amazing that this project progressed as far as it did and was then dropped so abruptly. We haven’t found any flaws in our virtual testing. If we misalign the layers as we suspect they did, then our results are similar to theirs.”

“So how long would it take to confirm your theory?” asked Sally.

“We think two weeks at the most,” replied Marléne.

“Probability of success?”

“Fifty – fifty,” replied several.

Then Gillian added, “Intuitively, we think it has the potential for a home run.”

“We don’t see any downside, so go for it,” said the Command Staff.

Siobhan asked, “So what are your next steps?”

“While we are confirming the virtual testing, we will build two prototypes identical to that in the documentation, with a few corrections. One correction is to ensure the alignment pins of each layer are significantly different. In parallel with this, the design and virtual prototyping will be progressing on the unit. In moving forward we will utilize the design information and our insight into how the current systems work. We plan to incorporate changes that weren’t known until after work stopped on the 3D chip development. These utilize new materials and current manufacturing technology. One concern we have is managing the heat, but we won’t know the answer to that until we build a couple.

“For the program side, our intention is to fully review what was completed for this processor. We want to compare it to what the AIs currently have and the areas where we know there are problems. Our goal is to have the program ready to test using the new processor once its functions are validated.”

Marléne said, “By combining the 3D array processor with the more powerful programming instruction logic, we intend to create a new AI. Its ability to handle multiple inputs should be considerably better than Star can currently do, while using less power. It will also likely be smaller. The last two are side benefits, if true, however there is nothing to indicate the opposite impact.”

Li Min said, “The first stage has two parts. One is building a new AI container with all the needed interfaces. The second is to create a new OS. Creating this would be a joint effort between our programmers and the ship’s AIs. After testing functionality of new AI container, the new OS would be installed and evaluated. Testing would involve our programmers and ship AIs. We hope this will avoid creating any decision tree paradoxes. After establishing that everything was functioning correctly, we would copy Star to the new container. After a cold start and system check, ship interfaces would be transferred to the new container. At this point, Star’s functions would be from the new container. During this time, we would need the Command Staff’s support as we verify how well Star functions using the new system. The departmental AIs will assist us in the evaluation.

Once Star’s functions in the new container are verified, the old container will be powered down. The old container will be refurbished, by replacing the old processor with the new design and updating all of its interfaces with the ship. We aren’t sure how we will handle the memory banks. The difficulty is that they are slower than the new version so mixing the two versions is a potential issue. Our preference is to change them, but we are waiting on Star’s assessment before making a decision.

Once Star’s original system is rebuilt, her functions will be transferred back to the original container then tested. The container that temporally held Star could then be configured as the primary AI for another entity, perhaps the Mars colony. One of the original programming goals was for the AIs to have a stronger positive symbiotic relationship with their biological counterparts. We are well aware of the potential for creating a ‘HAL’ like entity. Currently we think this can be avoided by ensuring a strong link between the crew and AIs. A full telepathic link between AIs and their primary crew contacts would greatly benefit them in other ways.”

“Nice job,” said Alison. “Star, have you and the other AIs reviewed this design.”