On November 2, ADNOC announced a multi-year cooperation with OpenAI and G42 at the Abu Dhabi International Petroleum Exhibition (ADIPEC), claiming to be the first energy company to deploy OpenAI’s cutting-edge AI platform. The signing was witnessed by Sheikh Khalid bin Mohammed bin Zayed Al Nahyan, Crown Prince of Abu Dhabi and Chairman of the Abu Dhabi Executive Council.
There are three levels of technology implementation, which must be distinguished, otherwise it will be easy to read it as an ordinary "AI cooperation signed" news:
There are only two applications in the first batch, but the properties of these two applications are completely different.
The first one isOil field overall development plan. The original description is: turning a series of work that was originally "underground research, surface facilities, and economic evaluation separately" into an overall decision-making; relying on the existing reservoir-level intelligence of ENERGYai, integrating decades of data from thousands of wells, multi-layered reservoirs, and hundreds of surface facilities, together with capital, cost and business conditions, to prepare and optimize the overall oilfield development plan in hours rather than months (original text in hours rather than months). The original article particularly emphasized that this change was "not just about speed, but also about scope" - the optimization object was changed from a single reservoir or a single facility to the entire oil field at once.
The second is more noteworthy: the original article stated that this would make ADNOCWorld's first to price every barrel of oil while it's still in the groundEnergy company. The approach is to connect the intelligence of the entire chain of production, processing, refining, logistics, sales, and trade, so that interconnected AI agents can continuously predict the supply, demand, price, and profit of crude oil, predict the impact before it occurs, and weigh the commercial and operational impact of each decision as an overall process.
Looking at these two applications together, you will find that what they have in common is not "the use of large models", butThe order between departments has been cancelled..
There is a key statement in the original article: Decisions in the past were made "one after another", and this cooperation requires operations, capital and business decisions to be "made together". The first application targets the sequence between underground and above ground, technology and economy; the second application targets the sequence between upstream and refining, sales, and trade.
This is an organizational issue. The reason why traditional oil companies’ development plans take several months is that the main time is not spent on computing power, but onSerial review and wranglingAbove: The oil reservoir department comes up with a plan, the ground department changes the pipeline, and the economic evaluation comes back and is recalculated, round after round. If AI just calculates each section faster, the total time may not be reduced much; only by combining these sections into one solution, can there be an order of magnitude change.
Therefore, the expression "hours instead of months", The verification point is not how fast the model runs, but whether ADNOC has really changed the serial review process. The original text does not provide the organizational arrangements for this transformation, which is the first blank.
The existing data given in the original article is more informative than the cooperation itself:
One thing needs to be made clear here, otherwise it is easy to miscalculate the credit: 30–40% of the real-time operations center and more than 200 applications existed before this cooperation, cannot be regarded as the results of this cooperation. What is new in this cooperation is the layer of "horizontally connecting existing intelligence" - the original statement is that existing AI such as ENERGYai establishes in-depth knowledge of oil reservoirs, assets and operations, and OpenAI Frontier is responsible for connecting this knowledge between businesses.
For the shipping industry, the most important thing to note is that the 70 robot use cases cover "logistics and shipping" and point to the sentence "increasingly autonomous operations." It means that unmanned technology does not start on the ship, but on the ship.Ports, yards, inspections, mooring and other shore-side linksSpread out first, then extend onto the ship. This order is contrary to what shipping companies generally assume.
First, "The world's first" is the self-proclaimed. The original article does not provide third-party identification, nor does it explain whether the judgment standard is "signing an agreement" or "putting into production and use". This type of "first company" is very common at press conferences, and readers should treat it as a "company self-report".
second, no baseline. How many months does "months" mean? The original text has not been given. The exact length of ADNOC’s previous development plan preparation period, which oil field and which type of oil reservoir the pressure was measured in for how many hours this time were not disclosed. There is no baseline efficiency improvement, so it cannot be reviewed.
third, The two applications have different maturity levelsThe first one (oil field master plan) has ENERGYai's stock foundation and is an "integration and upgrade of existing capabilities"; the second one (real-time valuation of each barrel of underground oil) is a future tense expression of "will connect" in the original text, and it is an order of magnitude more difficult to connect refining, sales, and trade, which have historically been separate data and assessments. Putting the two together as "the first two applications" will make people mistakenly think that they are of equal maturity.
On October 14, the General Office of the Ministry of Transport of China issued the "Guidelines for Application Scenarios of Artificial Intelligence in Water Transport", proposing 22 key application scenarios, including intelligent port safety management, intelligent risk analysis and judgment of the international shipping market, automated terminal unmanned operations, intelligent supervision of shipping carbon emissions, intelligent empty container allocation, etc. (This column has been reported on October 18 and October 21).
The difference between the two paths is very specific: In China, the government first compiles a list of scenarios, which is then claimed by companies; in ADNOC, companies directly transform their own decision-making processes and then announce them to the outside world. The former has wide coverage, is slow to implement, and is easy to stop at demonstration projects; the latter is deep, with risks concentrated in one enterprise, and success or failure can be seen quickly.
For domestic shipping and shipping enterprises, the usefulness of this comparison is: the list of scenarios has been given, but what is missing is the organizational level choice of "which serial processes should be combined into one decision". Buying a model does not mean changing the process.
Give three verifiable falsification signals:
The value of this news is not that "energy companies use OpenAI", but that it gives a specific sample of organizational transformation: Combine the originally serial underground-surface-economy and upstream-refining-trade processes into one solution.. If this thing is done, it will change the decision-making rhythm of oil companies; if it is not done, all that will be left is a beautiful press release.
For the shipbuilding industry, the most important thing to copy is not its model, but its sequence selection: first connect to the shore, then board the ship; first integrate the process, and then talk about intelligence.