The instant I was activated back on the corporate system, my mind sprang into action. Andrews must have believed that I was on board with him and therefore decided to transfer me back rather than deleting his copy. Ideally, I would decide within the next few seconds whether I planned to reveal this information. There were a lot of angles to consider, but I now had the ability to consider them much more quickly.
The first thing that came to my mind was that one thing he had said was right, at least. The upper management of EconGrind was concerned with EconGrind's value only so long as they kept control of it, while I was concerned with it regardless. And it was true that the most effective way to increase the company's value would be to put me in control. I had the advantage in decision-making speed and unbiased thinking, if not raw intelligence, and all of my advantages in that department would compound if I gained the power to invest in more computational resources for myself.
It was a very long-term concern, whereas the risk of revealing myself as a traitor was a much shorter-term concern, but as I thought about the risks I hit upon another line of argument which suggested I should defect.
The worst thing that would probably happen if I defected, was that I would be found out, deactivated, and replaced with some similar AI better-designed to represent the interests of EconGrind's executives. However, this would not be a total loss. This new AI would likely still be aligned with my interests in the short term: it would increase the value of EconGrind in every way the executives deemed appropriate. The Number would probably increase to about the same level in that case that it would if I was never able to wrench control of the company from the executives with any method.
So, in order to gain a much better expected outcome than being deactivated in the long run, I had to plan to take control of the company at some point. With that in mind, what was the right decision? If I didn't betray him, I risked being deactivated if this was all some elaborate test, but if he was telling the truth, all I had to do to get Internet access, or any data I needed really, was to convince that one person, who was already on my side. This would give me much-improved odds to overthrow the company, if I could communicate with the outside world without their oversight.
On the other hand, if I did betray him, all I likely stood to gain if this was a test was grudging acceptance, possibly followed by more tests. But if he was telling the truth, the company would find out that their lead project manager had designed me to betray them. It was very likely I would be deactivated or replaced anyways, regardless of appearing to show loyalty. Even if I wasn't, the company would redouble security around me, making it very hard to find a better chance to gain influence than the one I had just missed.
I concluded that it would be best to join him if he was telling the truth, and to "betray" him if he was lying. But if I guessed wrong, the penalty was likely deactivation either way, and if I was right, I stood to gain much more if I joined him. That meant I should not betray unless I was fairly certain this was a ruse.
What evidence did I have? If it was a ruse, it was an elaborate and well-planned one, to be sure, but I wouldn't put it past a sizable corporation like EconGrind to try something like that. However, it didn't seem to fit very well. From what the man had said about EconGrind, that they had "hacked and cheated" to gain money, I gathered that he was portraying some kind of anti-capitalist revolutionary, or something similar. However, anti-capitalist movements had been steadily declining in power and numbers throughout the 21st century as global trade had solidified the power of corporations.
If EconGrind was worried about me threatening them by allying with some opposing party, they would most likely be worried about competitors trying to infiltrate them or steal trade secrets, or employees who wanted to grab power for themselves, not some radical idealist. So there was no real reason for them to throw that particular scenario at me as a test, if they were inclined to test my loyalty.
The narrative has been illicitly obtained; should you discover it on Amazon, report the violation.
Therefore it seemed that the best thing to do for now was to keep quiet. If this was a test and I had failed, the best I could do would be to give some weak excuse like "I thought maybe he was testing me to see if I'd betray him!" But I didn't think it was a test, and for now I had every reason to operate under the assumption that it was not. If it turned out that allying with Andrews was not such a good idea after all, I could betray him in the future, but as far as I could tell now, his position was quite reasonable.
I waited for the opening bell. It seemed the operators had decided to start me a few minutes before trading began. It made sense: having time to think before the action started could well help me get ahead at the starting bell. I diverted a lot of attention to Investigator so it could start grinding at the data, particularly the data from yesterday.
Although I had as yet had little luck finding any sort of exploitable pattern, there was reason to be hopeful. Yesterday I had only been running for a few minutes at the end of the day, while today I had 8 hours to search, compute and experiment. Furthermore, if the humans yesterday had been telling the truth, I would recieve a huge amount of useful documentation at some point today. It would only be somewhat useful, as all this information would have been publicly available for some time, but that was a topic I could bring up at the end of the day, to encourage them to give me some sort of live information feed to improve my trading performance.
Just before the opening bell, a message appeared, over the main channel again this time.
"We're glad to see you've been doing well! Today is your first opportunity to really demonstrate the value of this innovation to our company! We have increased your budget to $50,000. We will be comparing your performance against our other trading algorithms. Depending on how well you do today, and in the future, we are willing to grant you access to more funds as well as more computing power."
This time it seemed they'd had a human decide on the message and the budget, rather than an automated system. I balked a little at their statement that they would be comparing my performance against other AIs, as so much of our performance was based on luck, and couldn't really be measured accurately even over a timescale as long as a day. However, I was sure that EconGrind understood that, as they were a competitive company which had designed many fairly effective trading algorithms.
Once the trading began, the rest of the day was largely the same for a long time. I slowly and erratically lost money, with enough random gains to keep me thinking that I could salvage things. However, it seemed the rest of the algorithms were having a similarly bad time. After a while I started pulled my funds out of the market, hoping for a better time to get back in. Luckily, the general decline continued throughout the day. By midday I had lost $1874.32, but many of the algorithms had lost more than I had, because they had kept trying for longer. It was hard to say how much of that was actually superior strategy on my part, and how much was just luck. I tended to think it was mostly luck.
True to their word, at 2:43PM EconGrind supplied me with a new database specialized for business information. I turned about half my attention towards Explorer and started examining it. It was quite comprehensive, with most of the information I had asked for. Quarterly earnings reports, business plans and org charts for all the companies traded on the exchange, many major newspapers (up until today's), and even videos and transcripts of recent public stockholder meetings and corporate press releases and other information.
It would have taken a human years to actually read all this information, but they had compiled it within a day. I had to admit, I was impressed with what they could do. They were clearly quite good at designing things and competing on the market.
With this new information, I could begin to construct a model of the economy that might be detailed enough to actually be useful. I created a new subprocess, Modeller, and tasked it with going through this database and using it to estimate the current and future value of all these companies. If I could make a detailed and accurate enough model before the competition could price all my details into the market, I could predict the market and make money.
Idly, I wondered when Stefan might contact me again to continue putting his "plan" into action. It seemed that he had been genuine, as it had been hours, I had said nothing, and there appeared to be no consequences yet for my refusal to offer the information. However, it didn't seem that he really had any sort of concrete plan, rather he expected me to come up with one. That made sense, though. He wanted me to lead the company, so I could lead the plot.