The billionth digit of pi is 9.

Full current name: Factoid Intelligence Sphere, also known as the Factual Intelligence Sphere or simply the Fact Sphere, though he personally prefers Craig in informal interactions. His original name was the Factual Intelligence Sphere, though it was changed to Factoid during later development. 

Craig was one of the earlier projects in AI development in his 'verse, and one of the later models Aperture scrapped in favour of burning actual personalities onto discs. The first core series - the Emotional Simulation Core, the Logical Intelligence Core, the Learning Development Core, and the Moral Attunement Core - came before Craig; the first series of artificial intelligence attempts, though these were deemed unsuccessful and instead became personality aspect cores specifically altered to enhance and control GLaDOS' personality (before Cave had the idea of just putting a real person’s personality in there). 

Taking cues first and foremost from the files of the original Logical Intelligence Core, the Factoid Intelligence Sphere was an experiment in developing an actual artificial personality as much as an experiment in morality coding. While Craig's dialogue still consisted primarily of listing pre-made lines, he also had the capability to observe and learn, and comment on the environment and situation around him. Once he managed to identify what the researchers around him were doing and comment on it, and even carry basic conversation, he was determined to be a successful AI. 

The morality aspect was harder - once he'd become relatively self-aware, he was quite content to simply repeat facts rather than hold full-fledged conversations, and would even argue to the researchers that as a Fact Sphere (as they'd taken to calling him for short) it was unnecessary for him to hold subjective opinions. As long as he could respond with factual statements and observations, though, he could still hold conversations with humans, and that was the most important thing. 

Eventually, they did manage to graft morality programs onto him - by adding emotional simulation programs first. For a brief while, Craig was at his most complete; his simulated emotions were restricted to some extent, but he held subjective opinions and finally appeared to have a human-like personality. 

The scientists could have stopped there, but they didn't. They wanted to be able to debate with him, to really test him on morality, subjective truth, and lies. This would be their downfall - they updated his files with upgraded emotion levels, and set his morality software to Auto, letting his personality, emotions, and arguments determine the morality output for a time; in addition, they added a batch of mostly true, mostly false, and outright false statements to his databank. In a heated argument with the scientists, he had argued one of the entirely false statements as truth - until one of the scientists triumphantly admitted that they'd added lies to his databank, virtually indistiguishable from the truths from Craig's access perspective. This had the unfortunate effect of overblowing his anger output, as well as registering as an unbelievable paradox - as a result, he overheated in seconds, leading to a hard reset of his system. 

When Craig woke up, recharged and rebooted, the paradoxical feeling lingered. While the exact "fact" in question had been deleted, the knowledge that he still likely had infactual facts listed as truths in his system led to a hard reversal of his personality, and he reverted immediately to a minimum of subjectability, refusing to hold meaningful conversations. Instead he simply generated streams of familiar facts, locking away all recent database additions. 

As an AI, this rendered him more or less useless in the opinion of the researchers, and he was sent to the corrupted core bin as a failed project. Separate from Craig's development, humanoid androids were being developed in a neighboring facility, and a few prototypes were delivered to Craig's facility. Ultimately they did not become the primary testing bots and AI bots, as it was deemed that they were less than optimal in terms of balance and performance, but they remained, empty in the laboratory stock rooms. 

The events of Portal 1 and 2 took place before Craig was brought out of the bin, and in that time, he had been largely dormant. Years and years had gone in the presence of only two other active spheres - the Space Sphere and the Adventure Sphere - and after the first few months, they had largely shut up together in a mutual attempt at conserving battery lifetime, and because they had fiercely bullied each other in the bin, developing a barely restrained animosity towards each other. Confronted with GLaDOS, Chell, and Wheatley, all three had returned to talking incessantly - even Craig had become more and more subjective over the course of a few minutes, accidentally unlocking his false facts in the process as well. Then the portal to space had opened - and both the Adventure Sphere and the Space Sphere had been sucked into space, alongside Wheatley. Only Craig remained with GLaDOS, and was promptly put to work once he'd been fitted into a spare android body.