OpenAI is no longer the biggest player in the AI market: enter Elon Musk’s Grok. The unconventionally named language generative is said to be based on Douglas Adams’ science fiction franchise the ‘Hitchhiker’s Guide to the Galaxy’. Musk is of course, famous for his ambitious claims when it comes to introducing new products into the market, and Grok is no exception. According to a blog post by the developers of the technology, Grok can “answer almost anything and, far harder, even suggest what questions to ask!” Coined by American science fiction writer Robert A. Heinlein, grok is a verb that means ‘to understand something intuitively or by empathy’.
Grok is still in its early testing stages and is not yet available on the open market – although it will ultimately be released to subscribers to X’s (formerly Twitter) Premium Service. According to the tech tycoon, Grok apparently has access to user posts on X, also currently owned by Elon Musk. According to him, Grok is “ based and loves sarcasm”. The word ‘based’ used to be slang, meaning ‘addicted to crack cocaine’. The efforts to reclaim the word by rapper Lil B has resulted in the word being used to refer to being authentically yourself. Grok apparently ‘speaks’ in a sarcastic manner, examples of which Musk has already posted on his personal account.
Grok’s access to Twitter is believed to be a massive advantage over other models on the market, as it allows the tool to access real-time information as it is being generated online. Other tools depend largely on older archives of the internet to draw on in their answers.
Musk’s AI company xAI is behind the development of the tool and staff describes the software as having a bit of a rebellious streak – something that many will have an interesting time exploring. Cynics are already expecting that the chatbot’s sense of humour will also have interesting legal ramifications going forward. The AI tool will also reportedly address NSFW queries, that other chatbots on the market – with good reason – shy away from.
All AI chatbots have a language model that provides the foundations for all its interactions with humans, and Grok is no different. Grok’s language model is called Grok-1, which is said to surpass GPT-3.5 in some respects. For example, Grok-1 is described as being capable of solving middle-school math problems – Grok passed a Hungarian high school maths exam with a ‘C’ grade. GPT 3.5 is the model that is available for free in the freely available version of ChatGPT. However, Grok-1 is apparently yet to match GPT-4. According to xAI, this is because GPT-4 is trained on a far larger amount of data and computer resources than Grok-1 currently has access to. GPT-4 is currently described as showing ‘human-level performance’, on professional benchmarks – including the US bar exam. According to Musk, Grok also rivals Meta’s LLaMA 2 and DeepMind co-founder Mustafa Suleyman’s Inflection.
The generative AI industry has seen explosive success in 2023, raising billions of dollars in investor interest. Some even warn that the pouring of investor funds is contributing towards the creation of a tech bubble, as the market is still to settle on how the technology can be commercialised for the market. The speed at which the technology is seeing development is also mind-boggling – ChatGPT 3.5 launched in November just last year, which means that it has had a full year on the market to train to reach the capacity it has today. Grok, which started development in April this year, already rivals it as an equal.
Musk is also frequently cited as warning people of the dangerous capabilities of AI technology. Just a few weeks ago, Musk had described AI as a ‘threat to humanity’ just a few weeks prior to unveiling Grok, at the global AI safety summit held in England. Speaking at the event, he went on to say “…for the first time, we have a situation where there’s something that is going to be far smarter than the smartest human. So, you know, we’re not stronger or faster than other creatures, but we are more intelligent. And here we are, for the first time really in human history, with something that’s going to be far more intelligent than us”, adding that while it was not clear whether humanity will really be capable of controlling the technology, “we can aspire to guide it in a direction that’s beneficial to humanity”.
Musk is also a co-founder of OpenAI and has even made claims that he is the reason that OpenAI exists at all. He pulled out of investing in the company following concerns about conflicts of interest with his other companies and disagreements with the developers about the speed at which the technology was being developed, suggesting that the team wasn’t paying enough attention to ‘safe’ AI development. Musk even signed an open letter earlier this year that called for a six-month break in AI development until appropriate safeguards were established. This did not achieve the intended result, which was a result that Musk himself was aware likely to happen. In his own words, signing the letter was his way of ‘going on the record’ as having voiced his perspective – or having warned the public earlier. After signing the open letter in March, Musk went on to incorporate his rival AI firm xAI in April of this year.
In July, Musk justified his decision to proceed with AI development, due to a pause that seemed unrealistic under current conditions. It is unclear how long Grok will remain in product testing: according to media reports, it is currently training its technology via software group Oracle’s cloud computing platform. X’s Premium+ users can only apply to test Grok out for themselves after this testing, although the company is also reportedly entertaining beta testers.
(Theruni Liyanage)